Search

Lymphoma is a group of malignant tumors originated in the lymphatic system. Automatic and accurate lymphoma segmentation in PET/CT volumes is critical yet challenging in the clinical practice. Recently, UNet-like architectures are widely used for medical image segmentation. The pure UNet-like architectures model the spatial correlation between the feature maps very well, whereas they discard the critical temporal correlation. Some prior work combines UNet with recurrent neural networks (RNNs) to utilize the spatial and temporal correlation simultaneously. However, it is inconvenient to incorporate some advanced techniques for UNet to RNNs, which hampers their further improvements. In this paper, we propose a recurrent dense siamese decoder architecture, which simulates RNNs and can densely utilize the spatial-temporal correlation between the decoder feature maps following a UNet approach. We combine it with a modified hyper dense encoder. Therefore, the proposed model is a UNet with a hyper dense encoder and a recurrent dense siamese decoder (HD-RDS-UNet). To stabilize the training process, we propose a weighted Dice loss with stable gradient and self-adaptive parameters. We perform patient-independent fivefold cross-validation on 3D volumes collected from whole-body PET/CT scans of patients with lymphomas. The experimental results show that the volume-wise average Dice score and sensitivity are 85.58% and 94.63%, respectively. The patient-wise average Dice score and sensitivity are 85.85% and 95.01%, respectively. The different configurations of HD-RDS-UNet consistently show superiority in the performance comparison. Besides, a trained HD-RDS-UNet can be easily pruned, resulting in significantly reduced inference time and memory usage, while keeping very good segmentation performance.

Recently, deep learning, especially convolutional neural networks, has achieved the remarkable results in natural image classification and segmentation. At the same time, in the field of medical image segmentation, researchers use deep learning techniques for tasks such as tumor segmentation, cell segmentation, and organ segmentation. Automatic tumor segmentation plays an important role in radiotherapy and clinical practice and is the basis for the implementation of follow-up treatment programs. This paper reviews the tumor segmentation methods based on deep learning in recent years. We first introduce the common medical image types and the evaluation criteria of segmentation results in tumor segmentation. Then, we review the tumor segmentation methods based on deep learning from technique view and tumor view, respectively. The technique view reviews the researches from the architecture of the deep learning and the tumor view reviews from the type of tumors.

As Internet-of-Things (IoT) applications today utilize many diverse devices to collect information, Bluetooth low energy (BLE), featuring low power and low cost, is one of the most promising wireless solutions. To meet the requirements of diverse IoT applications, the neighbor discovery process (NDP) in BLE networks requires low cost and low latency, which is one of the most challenging tasks in supporting such a large number of BLE devices. Since the choice of BLE parameters is essential for achieving the required performance of BLE NDP, many performance analysis models have been proposed, aiming to provide guidance for the parameter configuration in IoT applications. This article reviews and studies the BLE NDP models and BLE performance analysis models proposed over the period 2012–2020, considering the advantages and constraints in utilizing these models in IoT. The performance analysis models are divided into two categories: 1) probabilistic models and 2) Chinese reminder theory-based models. The model design, performance metrics, deployment constraints, analysis results, and use cases are discussed for research, development, and applications.

Voice assistant devices function as interaction gateways in the Internet-of-Things (IoT) smart home. By using voice assistants, users are able to control smart homes via speech commands. However, voice assistants introduce potential security risks and privacy disclosures. For example, malicious actors could impersonate genuine users to send smart home speech commands. Speaker/user verification thus becomes a critical issue for smart home security. This article proposes a secure method for speaker verification in IoT smart homes using millimeter-wave (mmWave) radar. Specifically, we utilize the radar to capture both vocal cord vibration (VCV) and lip motion (LM) as multimodal biometrics for identifying speakers. Traditional voice-based speaker verification methods are vulnerable to impostor attacks, such as replay attacks and voice synthesis attacks, that use recorded or imitated voice audio to spoof the system. Our approach is able to protect IoT smart homes from these attacks by continuously detecting the liveness of the user using mmWave sensing and deep learning techniques. Extensive experiments show that the proposed approach can achieve high verification accuracy and be more robust against imposter attacks.

Radio spectrum awareness, including understanding radio signal activities, is crucial for improving spectrum utilization, detecting security vulnerabilities, and supporting adaptive transmissions. Related tasks include spectrum sensing, identifying systems and terminals, and understanding various protocol layers. In this paper, we investigate various identification and classification tasks related to fading channel parameters, signal distortions, Medium Access Control (MAC) protocols, radio signal types, and cellular systems. Specifically, we utilize deep learning methods in those identification and classification tasks. Performance evaluations demonstrate the effectiveness of deep learning in those radio spectrum awareness tasks.

Ultrasound (US) is one of the most widely used imaging modalities in medical diagnosis. It has the advantages of real-time, low cost, noninvasive nature, and easy to operate. However, it also has the unique disadvantages of strong artifacts and noise and high dependence on the experience of doctors. In order to overcome the shortcomings of ultrasound diagnosis and help doctor improve the accuracy and efficiency of diagnosis, many computer aided diagnosis (CAD) systems have been developed. In recent years, deep learning has achieved great success in computer vision with its unique advantages. In the aspect of medical US image analysis, deep learning has also been exploited for its great potential and more and more researchers apply it to CAD systems. In this paper, we first introduce the deep learning models commonly used in medical US image analysis; Second, we review the data preprocessing methods of medical US images, including data augmentation, denoising, and enhancement; Finally, we analyze the applications of deep learning in medical US imaging tasks (such as image classification, object detection, and image reconstruction).

The current development of 5G technology is flourishing with widespread deployment across the world at a rapid pace. However, there is still a demand concerning 5G research for service and performance improvement. Research tasks include but are not limited to quality-of-service (QoS), energy efficiency, massive connectivity, reliable communications, and security. Due to the advancement of deep learning, numerous such research has utilized this technique. This article provides a comprehensive review of 5G communications research using deep learning. Specifically, we address the issues of low-density parity-check (LDPC) coding, massive multiple-input multiple-output (MIMO), non-orthogonal multiple access (NOMA), resource allocation, and security.

BACKGROUND: The novel coronavirus disease 2019 (COVID-19) constitutes a public health emergency globally. The number of infected people and deaths are proliferating every day, which is putting tremendous pressure on our social and healthcare system. Rapid detection of COVID-19 cases is a significant step to fight against this virus as well as release pressure off the healthcare system. OBJECTIVE: One of the critical factors behind the rapid spread of COVID-19 pandemic is a lengthy clinical testing time. The imaging tool, such as Chest X-ray (CXR), can speed up the identification process. Therefore, our objective is to develop an automated CAD system for the detection of COVID-19 samples from healthy and pneumonia cases using CXR images. METHODS: Due to the scarcity of the COVID-19 benchmark dataset, we have employed deep transfer learning techniques, where we examined 15 different pre-trained CNN models to find the most suitable one for this task. RESULTS: A total of 860 images (260 COVID-19 cases, 300 healthy and 300 pneumonia cases) have been employed to investigate the performance of the proposed algorithm, where 70% images of each class are accepted for training, 15% is used for validation, and rest is for testing. It is observed that the VGG19 obtains the highest classification accuracy of 89.3% with an average precision, recall, and F1 score of 0.90, 0.89, 0.90, respectively. CONCLUSION: This study demonstrates the effectiveness of deep transfer learning techniques for the identification of COVID-19 cases using CXR images.

Radio spectrum awareness is an important topic to overcome many challenges, such as spectrum utilization and sharing appearing with the development of technologies in wireless communications. Some practical tasks of radio spectrum awareness are related to signal detection and identification for improving the system’s reliability, efficiency, and security. Eye diagrams are essential for measurement tools used by engineers to simulate, evaluate and debug systems. Eye diagrams reflect many vital parameters for signal integrity degradation, such as timing jitter, crosstalk, and inter-symbol interference. Therefore, using an eye diagram containing a valuable feature from the system could be helpful for spectrum awareness tasks. In this paper, we use deep learning to study and identify classes within quadrature amplitude modulation using eye diagrams and explored related impacts to enable radio spectrum awareness. Our results show that deep learning neural networks capable of classifying quadrature amplitude modulation types with eye diagrams at presenting timing jitter and varying signal-to-noise ratios.

Vein contraction and venous compression typically caused by low temperature and excessive placement pressure can blur the captured finger vein images and severely impair the quality of extracted features. To improve the quality of captured finger vein image, this paper proposes a 26-layer generator network constrained by Neighbors-based Binary Patterns (NBP) texture loss to recover the clear image (guessing the original clear image). Firstly, by analyzing various types and degrees of blurred finger vein images captured in real application scenarios, a method to mathematically model the local and global blurriness using a pair of defocused and mean blur kernels is proposed. By iteratively and alternatively convoluting clear images with both kernels in a multi-scale window, a polymorphic blur training set is constructed for network training. Then, NBP texture loss is used for training the generator to enhance the deblurring ability of the network on images. Lastly, a novel network structure is proposed to retain more vein texture feature information, and two residual connections are added on both sides of the residual module of the 26-layer generator network to prevent degradation and overfitting. Theoretical analysis and simulation results show that the proposed neighbors-based binary-GAN (NB-GAN) can achieve better deblurring performance than the the-state-of-the-art approaches.

A behavior description helps analyze tiny objects, similar objects, objects with weak visual information, and objects with similar visual information. It plays a fundamental role in the identification and classification of dynamic objects in microscopic videos. To this end, we propose foldover features to describe the behavior of dynamic objects. Foldover is defined as: Each frame of an object's motion is superimposed on the same spatial plane in the spacetime order of the motion, the result of the superposition is the foldover of the object's motion. Foldover of an object contains temporal information, spatial information, behavior features and static features. Therefore, the features extracted based on the foldover of the object are the foldover features. In this work, we first generate foldover for each object in microscopic videos in X, Y and Z directions, respectively. Then, we extract foldover features from the X, Y and Z directions with statistical methods, respectively. The core content of this paper is to construct the foldovers and extract the foldover features. Through these two steps, the temporal information, spatial information, behavior features and static features of the object are enhanced and included in the foldover features. Furthermore, the description of the behavior of dynamic objects by the foldover features is strengthened. Finally, we use four different classifiers to test the effectiveness of the proposed foldover features. In the experiment, we use a microscopic sperm video dataset to evaluate the proposed foldover features, including three types of 1374 sperms, and obtain the highest classification accuracy of 96.5%.

Chronic obstructive pulmonary disease (COPD) is associated with morphologic abnormalities of airways with various patterns and severities. However, the way of effectively representing these abnormalities is lacking and whether these abnormalities enable to distinguish COPD from healthy controls is unknown. We propose to use deep convolutional neural network (CNN) to assess 3D lung airway tree from the perspective of computer vision, thereby constructing models of identifying COPD. After extracting airway trees from CT images, snapshots of their 3D visualizations are obtained from ventral, dorsal and isometric views. Using snapshots of each view, one deep CNN model is constructed and further optimized by Bayesian optimization algorithm to indentify COPD. The majority voting of three views presents the final prediction. Finally, the class-discriminative localization maps have been drawn to visually explain the CNNs' decisions. The models trained with single view (ventral, dorsal and isometric) of colorful snapshots present the similar accuracy (ACC) (86.8%, 87.5% and 86.7%) and the model after voting achieves the ACC of 88.2%. The ACC of the final voting model using gray and binary snapshots achieves 88.6% and 86.4%, respectively. Our specially designed CNNs outperform the typical off-the-shelf CNNs and the pre-trained CNNs with fine tuning. The class-discriminative regions of COPD are mainly located at central airways; however, regions in HC are scattering and located at peripheral airways. It is feasible to identify COPD using snapshots of 3D lung airway tree extracted from CT images via deep CNN. The CNNs can represent the abnormalities of airway tree in COPD and make accurate CT-based diagnosis of COPD.

Computer-Aided Diagnosis (CAD) can improve the accuracy of diagnosis effectively, reduce the rate of misdiagnosis, and provide the support for the valid decision. In clinical applications, high requirements are often imposed on the execution speed and accuracy of CAD systems. The classifier is regarded as the core of the CAD system, that is, the performance of the classifier will have a decisive influence on the operating affection of the CAD system. Extreme Learning Machine (ELM) is a fast learning algorithm using Single Hidden Layer Feedforward Neural Network (SLFN) structure. With its advantages in training speed, generalization performance and accuracy, ELM has draw attention in many research fields, including the development of CAD system. The applications of ELM in CAD are reviewed in this research. First, the mathematical model of ELM and framework of CAD system are briefly introduced. Then, the application of ELM in CAD is reviewed in detail, including the feature modeling method combined with ELM in CAD and the specific application of ELM. Finally, we summarized the current research status of CAD systems based on ELM, and the future work is prospected.

The texture edge continuity of a finger vein image is very important for the accuracy of feature extraction. However, the traditional inpainting methods which, without accurate texture constraints, are easy to cause the vein texture of the inpainted image to be blurred and break. A finger vein image inpainting method with Gabor texture constraints is proposed. The proposed method effectively protects the texture edge continuity of the inpainted image. Firstly, using the proposed vertical phase difference coding method, the Gabor texture feature matrix of the finger vein image, which can accurately describe the texture information, can be extracted from the Gabor filtering responses. Then, according to the local texture continuity of the finger vein image, the known pixels, which have different texture orientations with the center pixel in the patch, are filtered out using the Gabor texture constraining mechanism during the inpainting process. The proposed method eliminates irrelevant information interference in the inpainting process and has a more precise texture propagation. Simulation experiments of artificially synthetic images and acquired images show that the finger vein images inpainted by the proposed method have better texture continuity and higher image quality than the traditional methods which do not have accurate texture constraints. The proposed method improves the recognition performance of the finger vein identification system with the acquired damaged images.

In recent years, researches are concentrating on the effectiveness of Transfer Learning (TL) and Ensemble Learning (EL) techniques in cervical histopathology image analysis. However, there have been very few investigations that have described the stages of differentiation of cervical histopathological images. Therefore, in this article, we propose an Ensembled Transfer Learning (ETL) framework to classify well, moderate and poorly differentiated cervical histopathological images. First of all, we have developed Inception-V3, Xception, VGG-16, and Resnet-50 based TL structures. Then, to enhance the classification performance, a weighted voting based EL strategy is introduced. After that, to evaluate the proposed algorithm, a dataset consisting of 307 images, stained by three immunohistochemistry methods (AQP, HIF, and VEGF) is considered. In the experiment, we obtain the highest overall accuracy of 97.03% and 98.61% on AQP staining images and poor differentiation of VEGF staining images, individually. Finally, an additional experiment for classifying the benign cells from the malignant ones is carried out on the Herlev dataset and obtains an overall accuracy of 98.37%.

High-resolution CT images can clearly display anatomical structures but does not display functional information, while blurred PET images can display molecular and functional information of lesions but cannot clearly display morphological structures. Therefore, accurate PET-CT image registration, which is used for anatomical structure and functional information fusion, is a prerequisite for early stage cancer diagnosis. However, some hypermetabolic anatomical structures, such as brain and bladder, have low registration accuracy. To solve this problem, a 3D unsupervised network based on a metabolic constraint function and a multi-domain similarity measure (3D MC-MDS Net) is proposed for 3D PET-CT image registration. Specifically, a metabolic constraint model is established based on the standard uptake value (SUV) distribution of hypermetabolic regions such as brain, bladder, liver and heart, which reduces the excessive distortion on displacement vector field (DVF) caused by hypermetabolic anatomical structures in PET images. A DVF estimator is built based on 3D unsupervised convolutional neural networks and a spatial transformer is used for warping 3D PET images to 3D CT images. The generated registration results (PET image patches) and the original 3D CT image patches are used for calculating the spatial domain similarity (SD similarity) and frequency domain similarity (FD similarity). Finally, the loss function of the entire registration network is constructed by a weighted sum of SD similarity, FD similarity and a smoothness of DVF. A dataset consisted of 170 whole-body PET-CT images is used for registration accuracy evaluation. The proposed unsupervised registration network, 3D MC-MDS Net, can accurately learn the 3D registration model by using the training dataset with the metabolic constraint model, which significantly improves the registration accuracy.

This paper explores a low-cost portable visible light communication (VLC) system to support the increasing needs of lightweight mobile applications. VLC grows rapidly in the past decade for many applications (e.g., indoor data transmission, human sensing, and visual MIMO) due to its RF interference immunity and inherent high security. However, most existing VLC systems heavily rely on fixed infrastructures with less adaptability to emerging lightweight mobile applications. This work proposes Light Storage, a portable VLC system takes the advantage of commercial smartphone flashlights as the transmitter and a solar panel equipped with both data reception and energy harvesting modules as the receiver. Light Storage can achieve concurrent data transmission and energy harvesting from the visible light signals. It develops multi-level light intensity data modulation to increase data throughput and integrates the noise reduction functionality to allow portability under various lighting conditions. The system supports synchronization together with adaptive error correction to overcome both the linear and non-linear signal offsets caused by the low time-control ability from the commercial smartphones. Finally, the energy harvesting capability in Light Storage provides sufficient energy support for efficient short range communication. Light Storage is validated in both indoor and outdoor environments and can achieve over 98% data decoding accuracy, demonstrating the potential as an important alternative to support low-cost and portable short range communication.

A fusion image combining both anatomical and functional information obtained by registering medical images of two different modalities, Positron Emission Tomography (PET) and Computed Tomography (CT), is of great significance for medical image analysis and diagnosis. Medical image registration relies on similarity measure which is low between PET/CT image voxels and therefore PET/CT registration is a challenging task. To address this issue, this paper presents an unsupervised end-to-end method, DenseRegNet, for deformable 3D PET/CT image registration. The method consists of two stages: (1) predicting 3D displacement vector field (DVF); and (2) registering 3D image. In the 3D DVF prediction stage, a two-level similarity measure together with a deformation regularization is proposed as loss function to optimize network training.In the image registration stage, a resampler and a spatial transformer are utilized to obtain the registration results. In this paper, 663 pairs of Uptake Value (SUV) and Hounsfield Unit (Hu) patches of 106 patients, 227 pairs of SUV and Hu patches of 35 patients and 259 pairs of SUV and Hu patches of 35 patients are randomly selected as training, validation and test set, respectively. Normalized cross correlation (NCC), intersection over union (IoU) of liver bounding box and euclidean distance (ED) on landmark points are used to evaluate the registration results. Experiment results show that the proposed method, DenseRegNet, achieves the best results in terms of liver bounding box IoU and ED, and the second highest value of NCC. For a trained model, given a new pair of PET/CT images, the registration result can be obtained with only one forward calculation within 10 seconds. Through qualitative and quantitative analyses, we demonstrate that, compared with other deep learning registration models, the proposed DenseRegNet achieves improved results in the challenging deformable PET/CT registration task.

The main obstacle to image augmentation with Generative Adversarial Networks (GANs) is the need for a large amount of training data, but this is difficult for small datasets like Environmental Microorganisms (EMs). EM image analysis plays a vital role in environmental monitoring and protection, but it is often encountered with small datasets due to the difficulty of EM image collection. To this end, we propose an Enhanced Framework of GANs (EF-GANs) that combines geometric transformation methods and GANs for EM image augmentation. First of all, the color of an EM image has an insignificant impact on its class label, based on this fact, we perform color space augmentation to the original EM images. Secondly, we train EF-GANs with augmented EM images to generate utterly new EM images. Finally, we rotate the generated samples in various directions to obtain a more natural performance. In this study, we use VGG16 and ResNet50 networks to evaluate the proposed EF-GANs on 21 different types of EMs (420 EM images). It is observed that the average precision (AP) of VGG16 increases between 4.5% and 84.1% in 20 EM classes and one class remains unchanged. The AP of Resnet50 rises between 8.7% and 38.7% in 12 EM classes and reaches 100% in two EM classes. Furthermore, to reflect the generalization performance of EF-GANs, we employ an utterly new EM image dataset (630 EM images) to test the previous VGG16 networks. We select the VGG16 networks with original and optimal settings for all the EM classes, and for testing, optimal settings for a single EM class is considered. In the 20 of 21 one-vs-rest EM image classification tasks, the AP of VGG16 increases between 1.66% and 88.1%. The results demonstrate that the proposed EF-GANs can achieve outstanding performance in augmenting single EM images with high quality and resolution, thus, to improve the APs of EM image classification.

Non-negative matrix factorization (NMF) can be used in clustering, feature representation or blind source separation. Many NMF methods have been developed including least squares (LS) error, Kullback–Leibler (KL) divergence, Itakura–Saito (IS) divergence, Bregman-divergence, α -divergence, β -divergence, γ -divergence, convex, constrained, graph-regularized NMFs. The main contribution of this paper is to develop a framework to generalize the existing NMF methods and also provide new NMF methods. This paper constructs a general optimization model and develops a generic updating rule with a simple structure using a surrogate function, which possesses similar properties as the standard NMF methods. The experimental results, obtained using several standard databases, demonstrate the power of the work in which some new methods provide performance superior to that of the other existing methods.

Background Deep learning has made spectacular achievements in analysing natural images, but it faces challenges for medical applications partly due to inadequate images. Objective Aiming to classify malignant and benign pulmonary nodules using CT images, we explore different strategies to utilize the state-of-the-art deep convolutional neural networks (CNN). Methods Experiments are conducted using the Lung Image Database Consortium image collection (LIDC-IDRI), which is a public database containing 1018 cases. Three strategies are implemented including to 1) modify some state-of-the-art CNN architectures, 2) integrate different CNNs and 3) adopt transfer learning. Totally, 11 deep CNN models are compared using the same dataset. Results Study demonstrates that, for the model modification scheme, a concise CifarNet performs better than the other modified CNNs with more complex architectures, achieving an area under ROC curve of AUC = 0.90. Integrated CNN models do not significantly improve the classification performance, but the model complexity is reduced. Transfer learning outperforms the other two schemes and ResNet with fine-tuning leads to the best performance with an AUC = 0.94, as well as the sensitivity of 91% and an overall accuracy of 88%. Conclusions Model modification, model integration, and transfer learning can play important roles to identify and generate optimal deep CNN models in classifying pulmonary nodules based on CT images efficiently. Transfer learning is preferred when applying deep learning to medical imaging applications.

Accurate nucleus detection is of great importance in pathological image analyses and diagnoses, which is a critical prerequisite for tasks such as automated grading hepatocellular carcinoma (HCC) nuclei. This paper proposes an automated nucleus detection framework based on a stacked sparse autoencoder (SSAE) and a case-based postprocessing method (CPM) in a coarse-to-fine manner. SSAE, an unsupervised learning model, is first trained using image patches of breast cancer. Then, the transfer learning and sliding window techniques are applied to other cancers’ pathological images (HCC and colon cancer) to extract the high-level features of image patches via the trained SSAE. Subsequently, these high-level features are fed to a logistic regression classifier (LRC) to classify whether each image patch contains a complete nucleus in a coarse detection process. Finally, CPM is developed for refining the coarse detection results which removes false positive nuclei and locates adhesive or overlapped nuclei effectively. SSAE-CPM achieves an average nucleus detection accuracy of 0.8748 on HCC pathological images, which can accurately locate almost all nuclei on the pathological images with serious differentiation. In addition, our proposed detection framework is also evaluated on a public dataset of colon cancer, with a mean F1 score of 0.8355. Experimental results demonstrate the performance advantages of our proposed SSAE-CPM detection framework as compared with related work. While our detection framework is trained on the pathological images of breast cancer, it can be easily and effectively applied to nucleus detection tasks on other cancers without re-training.

In this paper, we study a multiuser uplink network consisting of one base station (BS), multiple users and one eavesdropper (E), where the users are intended to transmit their confidential messages to BS, while the eavesdropper attempts to tap their transmissions. To improve the transmission secrecy, we propose two friendly jammer selection-aided multiuser scheduling schemes, namely, the random jammer selection-aided multiuser scheduling (RJS-MUS) without knowing the eavesdropper’s channel state information (CSI) and the optimal jammer selection-aided multiuser scheduling (OJS-MUS), where the CSIs of eavesdropper are available. For comparison purposes, the conventional non-jammer selection-aided multiuser scheduling (NJS-MUS) scheme is considered as a benchmark. We derive exact and asymptotic closed-form secrecy outage probability expressions for the conventional NJS-MUS as well as proposed RJS-MUS and OJS-MUS schemes. The numerical results show that the proposed RJS-MUS and OJS-MUS schemes with equal power allocation between the selected friendly jammer and scheduled user perform worse than the conventional NJS-MUS approach in terms of the secrecy outage probability in the low signal-to-noise ratio (SNR) region. As the SNR increases, the secrecy outage performance of RJS-MUS and OJS-MUS schemes substantially improves, which is, in turn, better than that of conventional NJS-MUS approach. Moreover, the secrecy advantage of RJS-MUS and OJS-MUS over NJS-MUS becomes more significant with an increasing SNR. Also, it is shown that for both the RJS-MUS and OJS-MUS schemes, a better secrecy performance can be achieved through an optimal power allocation (OPA) between the scheduled user and friendly jammer. In addition, the proposed RJS-MUS and OJS-MUS schemes with OPA strictly outperform the conventional NJS-MUS approach in terms of the secrecy outage probability.

Deep learning (DL) is a new machine learning (ML) methodology that has found successful implementations in many application domains. However, its usage in communications systems has not been well explored. This paper investigates the use of the DL in modulation classification, which is a major task in many communications systems. The DL relies on a massive amount of data and, for research and applications, this can be easily available in communications systems. Furthermore, unlike the ML, the DL has the advantage of not requiring manual feature selections, which significantly reduces the task complexity in modulation classification. In this paper, we use two convolutional neural network (CNN)-based DL models, AlexNet and GoogLeNet. Specifically, we develop several methods to represent modulated signals in data formats with gridlike topologies for the CNN. The impacts of representation on classification performance are also analyzed. In addition, comparisons with traditional cumulant and ML-based algorithms are presented. Experimental results demonstrate the significant performance advantage and application feasibility of the DL-based approach for modulation classification.

Modulation classification is one of the key tasks for communications systems monitoring, management, and control for addressing technical issues, including spectrum awareness, adaptive transmissions, and interference avoidance. Recently, deep learning (DL)-based modulation classification has attracted significant attention due to its superiority in feature extraction and classification accuracy. In DL-based modulation classification, one major challenge is to preprocess a received signal and represent it in a proper format before feeding the signal into deep neural networks. This article provides a comprehensive survey of the state-of-the-art DL-based modulation classification algorithms, especially the techniques of signal representation and data preprocessing utilized in these algorithms. Since a received signal can be represented by either features, images, sequences, or a combination of them, existing algorithms of DL-based modulation classification can be categorized into four groups and are reviewed accordingly in this article. Furthermore, the advantages as well as disadvantages of each signal representation method are summarized and discussed.

Object identification or classification has found many applications, ranging from civilian to defense application scenarios and there is a rising need for a both effective and efficient identification approach. One of the common methods for this task is using a neural network. However, it could be very difficult for such a network to obtain accurate answers due to complex environments, especially when the data is of single modality. In this paper, we attempt to build a combined deep learning model which takes two distinct data modalities to help us achieve high accuracy multimodal classification systems. An experiment is conducted on Multimodal Unmanned Aerial Vehicle Dataset for Low Altitude Traffic Surveillance (AU-AIR) using both visual and sensor data. We compare our results between a model trained with visual data only and another combined model trained with both visual and sensor data. Improved object classification performance is observed when the multimodal method is applied.

Freezing of gait (FOG) has devastating consequences for patients with Parkinson's disease (PD), but the underlying pathophysiological mechanism is unclear. This was investigated in the present study by integrated structural and functional connectivity analyses of PD patients with or without FOG (PD FOG+ and PD FOG–, respectively) and healthy control (HC) subjects. We performed resting-state functional magnetic resonance imaging (fMRI) and diffusion tensor imaging of 24 PD FOG+ patients, 37 PD FOG– patients, and 24 HCs. Tract-based spatial statistics was applied to identify white matter (WM) abnormalities across the whole brain. Fractional anisotropy (FA) and mean diffusivity (MD) of abnormal WM areas were compared among groups, and correlations between these parameters and clinical severity as determined by FOG Questionnaire (FOGQ) score were analyzed. Voxel-mirrored homotopic connectivity (VMHC) was calculated to identify brain regions with abnormal interhemispheric connectivity. Structural and functional measures were integrated by calculating correlations between VMHC and FOGQ score and between FA, MD, and VMHC. The results showed that PD FOG+ and PD FOG– patients had decreased FA in the corpus callosum (CC), cingulum (hippocampus), and superior longitudinal fasciculus and increased MD in the CC, internal capsule, corona radiata, superior longitudinal fasciculus, and thalamus. PD FOG+ patients had more WM abnormalities than PD FOG– patients. FA and MD differed significantly among the splenium, body, and genu of the CC in all three groups (P < 0.05). The decreased FA in the CC was positively correlated with FOGQ score. PD FOG+ patients showed decreased VMHC in the post-central gyrus (PCG), pre-central gyrus, and parietal inferior margin. In PD FOG+ patients, VMHC in the PCG was negatively correlated with FOGQ score but positively correlated with FA in CC. Thus, FOG is associated with impaired interhemispheric brain connectivity measured by FA, MD, and VMHC, which are related to clinical FOG severity. These results demonstrate that integrating structural and functional MRI data can provide new insight into the pathophysiological mechanism of FOG in PD.

Microorganisms such as bacteria and fungi play essential roles in many application fields, like biotechnique, medical technique and industrial domain. Microorganism counting techniques are crucial in microorganism analysis, helping biologists and related researchers quantitatively analyze the microorganisms and calculate their characteristics, such as biomass concentration and biological activity. However, traditional microorganism manual counting methods, such as plate counting method, hemocytometry and turbidimetry, are time-consuming, subjective and need complex operations, which are difficult to be applied in large-scale applications. In order to improve this situation, image analysis is applied for microorganism counting since the 1980s, which consists of digital image processing, image segmentation, image classification and suchlike. Image analysis-based microorganism counting methods are efficient comparing with traditional plate counting methods. In this article, we have studied the development of microorganism counting methods using digital image analysis. Firstly, the microorganisms are grouped as bacteria and other microorganisms. Then, the related articles are summarized based on image segmentation methods. Each part of the article is reviewed by methodologies. Moreover, commonly used image processing methods for microorganism counting are summarized and analyzed to find common technological points. More than 144 papers are outlined in this article. In conclusion, this paper provides new ideas for the future development trend of microorganism counting, and provides systematic suggestions for implementing integrated microorganism counting systems in the future. Researchers in other fields can refer to the techniques analyzed in this paper.

Cervical cancer, one of the most common fatal cancers among women, can be prevented by regular screening to detect any precancerous lesions at early stages and treat them. Pap smear test is a widely performed screening technique for early detection of cervical cancer, whereas this manual screening method suffers from high false-positive results because of human errors. To improve the manual screening practice, machine learning (ML) and deep learning (DL) based computer-aided diagnostic (CAD) systems have been investigated widely to classify cervical pap cells. Most of the existing researches require pre-segmented images to obtain good classification results, whereas accurate cervical cell segmentation is challenging because of cell clustering. Some studies rely on handcrafted features, which cannot guarantee the classification stage's optimality. Moreover, DL provides poor performance for a multiclass classification task when there is an uneven distribution of data, which is prevalent in the cervical cell dataset. This investigation has addressed those limitations by proposing DeepCervix, a hybrid deep feature fusion (HDFF) technique based on DL to classify the cervical cells accurately. Our proposed method uses various DL models to capture more potential information to enhance classification performance. Our proposed HDFF method is tested on the publicly available SIPAKMED dataset and compared the performance with base DL models and the LF method. For the SIPAKMED dataset, we have obtained the state-of-the-art classification accuracy of 99.85%, 99.38%, and 99.14% for 2-class, 3-class, and 5-class classification. Moreover, our method is tested on the Herlev dataset and achieves an accuracy of 98.32% for binary class and 90.32% for 7-class classification., 12 pages, 8 figures, Published in Computers in Biology and Medicine

To recognize the epidermal growth factor receptor (EGFR) gene mutation status in lung adenocarcinoma (LADC) has become a prerequisite of deciding whether EGFR-tyrosine kinase inhibitor (EGFR-TKI) medicine can be used. Polymerase chain reaction assay or gene sequencing is for measuring EGFR status, however, the tissue samples by surgery or biopsy are required. We propose to develop deep learning models to recognize EGFR status by using radiomics features extracted from non-invasive CT images. Preoperative CT images, EGFR mutation status and clinical data have been collected in a cohort of 709 patients (the primary cohort) and an independent cohort of 205 patients. After 1,037 CT-based radiomics features are extracted from each lesion region, 784 discriminative features are selected for analysis and construct a feature mapping. One Squeeze-and-Excitation (SE) Convolutional Neural Network (SE-CNN) has been designed and trained to recognize EGFR status from the radiomics feature mapping. SE-CNN model is trained and validated by using 638 patients from the primary cohort, tested by using the rest 71 patients (the internal test cohort), and further tested by using the independent 205 patients (the external test cohort). Furthermore, SE-CNN model is compared with machine learning (ML) models using radiomics features, clinical features, and both features. EGFR(-) patients show the smaller age, higher odds of female, larger lesion volumes, and lower odds of subtype of acinar predominant adenocarcinoma (APA), compared with EGFR(+). The most discriminative features are for texture (614, 78.3%) and the features of first order of intensity (158, 20.1%) and the shape features (12, 1.5%) follow. SE-CNN model can recognize EGFR mutation status with an AUC of 0.910 and 0.841 for the internal and external test cohorts, respectively. It outperforms the CNN model without SE, the fine-tuned VGG16 and VGG19, three ML models, and the state-of-art models. Utilizing radiomics feature mapping extracted from non-invasive CT images, SE-CNN can precisely recognize EGFR mutation status of LADC patients. The proposed method combining radiomics features and deep leaning is superior to ML methods and can be expanded to other medical applications. The proposed SE-CNN model may help make decision on usage of EGFR-TKI medicine.

Background To assist doctors to diagnose mild cognitive impairment (MCI) and Alzheimer's disease (AD) early and accurately, convolutional neural networks based on structural magnetic resonance imaging (sMRI) images have been developed and shown excellent performance. However, they are still limited in their capacity in extracting discriminative features because of large sMRI image volumes yet small lesion regions and the small number of sMRI images. Methods We proposed a task-driven hierarchical attention network (THAN) taking advantage of the merits of patch-based and attention-based convolutional neural networks for MCI and AD diagnosis. THAN consists of an information sub-network and a hierarchical attention sub-network. In the information sub-network, an information map extractor, a patch-assistant module, and a mutual-boosting loss function are designed to generate a task-driven information map, which automatically highlights disease-related regions and their importance for final classification. In the hierarchical attention sub-network, a visual attention module and a semantic attention module are devised based on the information map to extract discriminative features for disease diagnosis. Results Extensive experiments were conducted for four classification tasks: MCI versus (vs.) normal controls (NC), AD vs. NC, AD vs. MCI, and AD vs. MCI vs. NC. Results demonstrated that THAN attained the accuracy of 81.6% for MCI vs. NC, 93.5% for AD vs. NC, 80.8% for AD vs. MCI, and 62.9% for AD vs. MCI vs. NC. It outperformed advanced attention-based and patch-based methods. Moreover, information maps generated by the information sub-network could highlight the potential biomarkers of MCI and AD, such as the hippocampus and ventricles. Furthermore, when the visual and semantic attention modules were combined, the performance of the four tasks was highly improved. Conclusions The information sub-network can automatically highlight the disease-related regions. The hierarchical attention sub-network can extract discriminative visual and semantic features. Through the two sub-networks, THAN fully exploits the visual and semantic features of disease-related regions and meanwhile considers global features of sMRI images, which finally facilitate the diagnosis of MCI and AD.

Type 2 diabetes mellitus (T2DM) leads to a higher risk of brain damage and adversely affects cognition. The underlying neural mechanism of T2DM-induced cognitive impairment (T2DM-CI) remains unclear. This study proposes to identify a small number of dysfunctional brain connections as imaging biomarkers, distinguishing between T2DM-CI, T2DM with normal cognition (T2DM-NC), and healthy controls (HC). We have recruited 22 T2DM-CI patients, 31 T2DM-NC patients, and 39 HCs. The structural Magnetic Resonance Imaging (MRI) and resting state fMRI images are acquired, and neuropsychological tests are carried out. Amplitude of low frequency fluctuations (ALFF) is analyzed to identify impaired brain regions implicated with T2DM and T2DM-CI. The functional network is built and all connections connected to impaired brain regions are selected. Subsequently, L1-norm regularized sparse canonical correlation analysis and sparse logistic regression are used to identify discriminative connections and Support Vector Machine is trained to realize three two-category classifications. It is found that single-digit dysfunctional connections predict T2DM and T2DM-CI. For T2DM-CI versus HC, T2DM-NC versus HC, and T2DM-CI versus T2DM-NC, the number of connections is 6, 7, and 5 and the area under curve (AUC) can reach 0.912, 0.901, and 0.861, respectively. The dysfunctional connection is mainly related to Default Model Network (DMN) and long-distance links. The strength of identified connections is significantly different among groups and correlated with cognitive assessment score (p < 0.05). Via ALFF analysis and further feature selection algorithms, a small number of dysfunctional brain connections can be identified to predict T2DM and T2DM-CI. These connections might be the imaging biomarkers of T2DM-CI and targets of intervention.

Gastric cancer is a common and deadly cancer in the world. The gold standard for the detection of gastric cancer is the histological examination by pathologists, where Gastric Histopathological Image Analysis (GHIA) contributes significant diagnostic information. The histopathological images of gastric cancer contain sufficient characterization information, which plays a crucial role in the diagnosis and treatment of gastric cancer. In order to improve the accuracy and objectivity of GHIA, Computer-Aided Diagnosis (CAD) has been widely used in histological image analysis of gastric cancer. In this review, the CAD technique on pathological images of gastric cancer is summarized. Firstly, the paper summarizes the image preprocessing methods, then introduces the methods of feature extraction, and then generalizes the existing segmentation and classification techniques. Finally, these techniques are systematically introduced and analyzed for the convenience of future researchers.

Malignant cerebral edema (MCE) after an ischemic stroke results in a poor outcome or death. Early prediction of MCE helps to identify subjects that could benefit from a surgical decompressive craniectomy. Net water uptake (NWU) in an ischemic lesion is a predictor of MCE; however, CT perfusion and lesion segmentation are required. This paper proposes a new Image Patch-based Net Water Uptake (IP-NWU) procedure that only uses non-enhanced admission CT and does not need lesion segmentation. IP-NWU is calculated by comparing the density of ischemic and contralateral normal patches selected from the middle cerebral artery (MCA) area using standard reference images. We also compared IP-NWU with the Segmented Region-based NWU (SR-NWU) procedure in which segmented ischemic regions from follow-up CT images are overlaid onto admission images. Furthermore, IP-NWU and its combination with imaging features are used to construct predictive models of MCE with a radiomics approach. In total, 116 patients with an MCA infarction (39 with MCE and 77 without MCE) were included in the study. IP-NWU was significantly higher for patients with MCE than those without MCE (p < 0.05). IP-NWU can predict MCE with an AUC of 0.86. There was no significant difference between IP-NWU and SR-NWU, nor between their predictive efficacy for MCE. The inter-reader and interoperation agreement of IP-NWU was exceptional according to the Intraclass Correlation Coefficient (ICC) analysis (inter-reader: ICC = 0.92; interoperation: ICC = 0.95). By combining IP-NWU with imaging features through a random forest classifier, the radiomics model achieved the highest AUC (0.96). In summary, IP-NWU and radiomics models that combine IP-NWU with imaging features can precisely predict MCE using only admission non-enhanced CT images scanned within 24 h from onset.

Secondary structure plays an important role in determining the function of non-coding RNAs. Hence, identifying RNA secondary structures is of great value to research. Computational prediction is a mainstream approach for predicting RNA secondary structure. Unfortunately, even though new methods have been proposed over the past 40 years, the performance of computational prediction methods has stagnated in the last decade. Recently, with the increasing availability of RNA structure data, new methods based on machine-learning technologies, especially deep learning, have alleviated the issue. In this review, we provide a comprehensive overview of RNA secondary structure prediction methods based on machine-learning technologies and a tabularized summary of the most important methods in this field. The current pending issues in the field of RNA secondary structure prediction and future trends are also discussed., 25 pages, 5 figures, 1 table

Accurate lymphoma segmentation on Positron Emission Tomography (PET) images is of great importance for medical diagnoses, such as for distinguishing benign and malignant. To this end, this paper proposes an adaptive weighting and scalable distance regularized level set evolution (AW-SDRLSE) method for delineating lymphoma boundaries on 2D PET slices. There are three important characteristics with respect to AW-SDRLSE: 1) A scalable distance regularization term is proposed and a parameter $q$ can control the contour's convergence rate and precision in theory. 2) A novel dynamic annular mask is proposed to calculate mean intensities of local interior and exterior regions and further define the region energy term. 3) As the level set method is sensitive to parameters, we thus propose an adaptive weighting strategy for the length and area energy terms using local region intensity and boundary direction information. AW-SDRLSE is evaluated on 90 cases of real PET data with a mean Dice coefficient of 0.8796. Comparative results demonstrate the accuracy and robustness of AW-SDRLSE as well as its performance advantages as compared with related level set methods. In addition, experimental results indicate that AW-SDRLSE can be a fine segmentation method for improving the lymphoma segmentation results obtained by deep learning (DL) methods significantly.

The ever-growing security issues in various mobile applications and smart devices create an urgent demand for a reliable and convenient user verification method. Traditional verification methods request users to provide their secrets (e.g., entering passwords and collecting fingerprints). We envision that the essential trend of user verification is to free users from active participation in the verification process. Toward this end, we propose a continuous user verification system, which re-uses the widely deployed WiFi infrastructure to capture the unique physiological characteristics rooted in user’s respiratory motions. Different from the existing continuous verification approaches, posing dependency on restricted scenarios/user behaviors (e.g., keystrokes and gaits), our system can be easily integrated into any WiFi infrastructure to provide non-intrusive continuous verification. Specifically, we extract the respiration-related signals from the channel state information (CSI) of WiFi. We then derive the user-specific respiratory features based on the waveform morphology analysis and fuzzy wavelet transformation of the respiration signals. Additionally, a deep learning based user verification scheme is developed to identify legitimate users accurately and detect the existence of spoofing attacks. Extensive experiments involving 20 participants demonstrate that the proposed system can robustly verify/identify users and detect spoofers under various types of attacks.

Wireless communications systems design and its performance depend on the wireless fading channels, which are often characterized using a Rician probability function. A Rician K-factor is used to describe the fading severity in a Rician fading channel and is used in the system design and performance evaluation. Therefore, the estimation of the Rician K-factor is important in wireless communications research and development. Traditionally, a Rician K-factor equation, the statistics of the instantaneous frequency of the received signal with a lookup table, or the James-Stein estimator with the maximum likelihood estimation is used for the K-factor estimation. In this paper, we explore the use of deep learning for K-factor estimation. Specifically, we use the convolutional neural network (CNN) to estimate the Rician K-factor from a waveform signal in a Rician channel. Numerical results demonstrate its good performance in estimating the K-factor of the Rician channel.

Spectrum awareness allows the understanding of the wireless systems environment and it gives engineers and designers better control in systems design and analysis. Phase noise is one of the characteristics of the channel distortion or device distortion, which causes transmission errors. In this paper, a deep learning network is utilized to study and identify different phase noise levels for quadrature phase shift keying (QPSK) signals. Our experiment results show that the deep learning neural network is capable of classifying a wide range of phase noise levels.

Making network nodes aware of the spectrum parameters can help to improve the spectrum utilization and network efficiency. To achieve such goals, machine learning (ML) and deep learning (DL) have been utilized to identify spectrum parameters, such as modulation formats, power levels, medium access control (MAC) protocols, etc. This paper explores MAC protocol identification using ML and DL in additive white Gaussian noise (AWGN) and Rayleigh fading environments. We transform the received signals into spectrogram and utilize convolutional neural networks (CNN) to recognize the MAC protocols. Experimentation results demonstrate the effectiveness in MAC protocol identification using ML and DL algorithms.

Spectrum awareness, including identifying different types of signals, is very important in a cellular system environment. In this paper, a neural network is utilized to identify 5G signals among different cellular communications signals, including Long-Term Evolution (LTE) and Universal Mobile Telecommunication Service (UMTS). We explore the use of deep learning in wireless communications systems. We consider the effects of training dataset size, features extracted, and channel fading in our study. Experiment results demonstrate the effectiveness of deep learning neural networks in identifying cellular system signals, including UMTS, LTE, and 5G.

Since the Internet of Things (IoT) applications today employ many different sensors to provide information, a large number of the Bluetooth low energy (BLE) devices will be developed as part of the IoT systems. The low-power and low-cost requirements of all BLE nodes are among the most challenging issues when supporting the neighbor discovery process (NDP) for such a large number of devices. Since the parameter settings are essential to achieve the required performance for the NDP, a parameter configuration method for neighbor discovery in BLE can be beneficial for determining some critical parameter metrics, such as the AdvInterval, ScanInterval, and ScanWindow. In this paper, we propose a parameter configuration scheme to balance the tradeoff between discovery latency and energy consumption. In the proposed scheme, the neighbor discovery latency and average energy consumption are expressed based on the Chinese Remainder Theory (CRT). With the expected primary performance, the parameters are configured accordingly using the parameter configuration algorithm. The experimental results show that the performance of the neighbor discovery varies with the parameter settings. Furthermore, two typical IoT applications are assessed in this paper. Compared with the simulation results, the proposed parameter configuration scheme can achieve high accuracy.

Accurate segmentation of bi-ventricle from cardiac magnetic resonance images (MRI) is a critical step in cardiac function analysis and disease diagnosis. Due to the morphological diversification of the heart and the factors of MRI itself, fully automated and concurrent bi-ventricle segmentation is a well-known challenge. In this paper, we propose cascaded conditional generative adversarial networks (C-cGANs) to divide the problem into two segmentation subtasks: binary segmentation for region of interest (ROI) extraction and bi-ventricle segmentation. In both subtasks, we adopt adversarial training that makes discriminator network to discriminate segmentation maps either from generator network or ground-truth which aims to detect and correct pixel-wise inconsistency between the sources of segmentation maps. For capturing more spatial information with multi-scale semantic features, in the generator network, we insert a multi-scale attention fusion (MSAF) module between the encoder and decoder paths. The experiment on ACDC 2017 dataset shows that the proposed model outperforms other state-of-the-art methods in most metrics. Moreover, we validate the generalization capability of this model on MS-CMRSeg 2019 and RVSC 2012 datasets without fine-tuning, and the results demonstrate the effectiveness and robustness of the proposed method for bi-ventricle segmentation.

CT screening has been proven to be effective for diagnosing lung cancer at its early manifestation in the form of pulmonary nodules, thus decreasing the mortality. However, the exponential increase of image data makes their accurate assessment a very challenging task given that the number of radiologists is limited and they have been overworked. Recently, numerous methods, especially ones based on deep learning with convolutional neural network (CNN), have been developed to automatically detect and classify pulmonary nodules in medical images. In this paper, we present a comprehensive analysis of these methods and their performances. First, we briefly introduce the fundamental knowledge of CNN as well as the reasons for their suitability to medical images analysis. Then, a brief description of various medical images datasets, as well as the environmental setup essential for facilitating lung nodule investigations with CNNs, is presented. Furthermore, comprehensive overviews of recent progress in pulmonary nodule analysis using CNNs are provided. Finally, existing challenges and promising directions for further improving the application of CNN to medical images analysis and pulmonary nodule assessment, in particular, are discussed. It is shown that CNNs have transformed greatly the early diagnosis and management of lung cancer. We believe that this review will provide all the medical research communities with the necessary knowledge to master the concept of CNN so as to utilize it for improving the overall human healthcare system.

A computer-aided diagnosis (CAD) system based on mammograms enables early breast cancer detection, diagnosis, and treatment. However, the accuracy of the existing CAD systems remains unsatisfactory. This paper explores a breast CAD method based on feature fusion with convolutional neural network (CNN) deep features. First, we propose a mass detection method based on CNN deep features and unsupervised extreme learning machine (ELM) clustering. Second, we build a feature set fusing deep features, morphological features, texture features, and density features. Third, an ELM classifier is developed using the fused feature set to classify benign and malignant breast masses. Extensive experiments demonstrate the accuracy and efficiency of our proposed mass detection and breast cancer classification method.

In the positron emission tomography/computed tomography (PET/CT) image diagnosis report, the semantic analysis of image findings section is an important part of the automatic diagnosis of medical image, which is an essential step for extracting keywords and abnormal sentences in the diagnostic report. To this end, this paper combines visibility attribute extraction network (VAE-Net) and bi-directional gated recurrent unit (BiGRU) into cascade networks to solve the tasks of attribute extraction and anomaly detection. First, a visibility attribute (VA) is defined to summary the vocabulary into 12 patterns based on the language characteristics in image findings. Second, a visibility attribute extraction network (VAE-Net) is developed to automatically extract VA from word embeddings, which is composed of residual convolutional neural network (residual CNN), BiGRU, and conditional random field (CRF). Finally, word embeddings and the corresponding VA are input into BiGRU and softmax to perform sentence-level anomaly detections. We evaluate the proposed method on a proprietary Chinese PET/CT diagnostic report dataset with an F1-score of 94.35% in the attribute extraction, an F1-score of 96.40% in sentence-level anomaly detection, and an F1-score of 96.77% in case-level anomaly detection. Besides, a publicity English national center for biotechnology information (NCBI) disease corpus dataset is used for externed validation with an F1-score of 95.81% in disease detection. The experimental results demonstrate the advantage of the proposed cascade networks as compared to other related methods.

Spectral reconstruction method based on narrow-band measurements has been demonstrated to achieve ultrafast spectroscopic imaging with high spatial and spectral resolution, in which multiple narrow-band images are collected by using several specific filters. Although commercially available filters can be employed in such method, filters with complex transmittance that are difficult to be fabricated typically show significant improvement in spectral reconstruction accuracy. In this study, a two dimensional programmable optical filter based on digital micromirror device (DMD) is proposed, in which its transmittance spectrum can be arbitrarily and quickly switched to realize complex transmittance. Furthermore, its flexible transmittance enables directly hardware-based spectral data post-processing, which can perform data acquisition and analysis simultaneously. Those have been evaluated by the diffuse reflectance spectra from normal and occluded skin flaps, as well as Raman spectra from live, apoptosis and necrosis leukemia cells. Our simulation results show that much higher spectral reconstruction accuracies can be achieved by the optimized filters with complex transmittance. Furthermore, the classification accuracy by using the proposed method is comparable to those achieved by conventional numerical methods. Therefore, based on the proposed programmable optical filter, fast spectroscopic imaging with high spatial and spectral resolution can be achieved for observing fast changing phenomena and even real-time target identification.

Various deep convolutional neural networks (CNNs) have been used to distinguish between benign and malignant pulmonary nodules using CT images. However, single learner usually presents unsatisfied performance due to limited hypothesis space, or falling into local minima, or wrong selection of hypothesis space. To tackle these issues, we propose to build ensemble learners through fusing multiple deep CNN learners for pulmonary nodules classification. CT image patches of 743 nodules are extracted from LIDC-IDRI database and utilized. First, eight deep CNN learners with different architectures are trained and evaluated by 10-fold cross-validation. Each nodule has eight predictions from the eight primary learners. Second, we fuse these eight predictions by the strategies of majority voting (VOT), averaging (AVE), or machine learning. Specifically, different machine learning algorithms including K-Nearest-Neighbor (KNN), Support Vector Machines (SVM), Naive Bayes (NB), Decision Trees (DT), Multi-layer Perceptron (MLP), Random Forests (RF), Gradient Boosting Regression Trees (GBRT) and Adaptive Boosting (AdaBoost) are implemented. Moreover, the correlation coefficients between the predictions of 10 ensemble learners are calculated, and the hierarchical clustering dendrogram is drawn. It is found that the ensemble learners achieve higher prediction accuracy (84.0% vs 81.7%) than single CNN learner. The overlap ratio among the 10 ensemble learners is much higher than that of the 8 primary learners (62.9% vs 33.2%). In addition, it is shown that ensemble learners are roughly divided into three categories: the first (SVM, MLP, GBRT and RF) achieves the best performance; the second (VOT and AVE) is better than the third (AdaBoost, DT, NB and KNN). VOT and AVE yield higher recall than the machine learning algorithms. These results indicate that ensemble learners based on multiple CNN learners can achieve better performances for pulmonary nodules classification using CT images and that preferred fusion strategies include SVM, MLP, GBRT and RF.

Liver cancer is one of the most common cancers. Liver tumor segmentation is one of the most important steps in treating liver cancer. Accurate tumor segmentation on computed tomography (CT) images is a challenging task due to the variation of the tumor’s shape, size, and location. To this end, this paper proposes a liver tumor segmentation method on CT volumes using multi-scale candidate generation method (MCG), 3D fractal residual network (3D FRN), and active contour model (ACM) in a coarse-to-fine manner. First, livers are segmented using 3D U-Net and then MCG is performed on these liver regions for obtaining tumor candidates (all superpixel blocks). Second, 3D FRN is proposed to further determine tumor regions, which is considered as coarse segmentation results. Finally, the ACM is used for tumor segmentation refinement. The proposed 3D MCG-FRN + ACM is trained using the 110 cases in the LiTS dataset and evaluated on a public liver tumor dataset of the 3DIRCADb with dice per case of 0.67. The experimentations and comparisons demonstrate the performance advantage of the 3D MCG-FRN + ACM compared to other segmentation methods.

Cervical cancer is a very common and fatal type of cancer in women. Cytopathology images are often used to screen for this cancer. Given that there is a possibility that many errors can occur during manual screening, a computer-aided diagnosis system based on deep learning has been developed. Deep learning methods require a fixed dimension of input images, but the dimensions of clinical medical images are inconsistent. The aspect ratios of the images suffer while resizing them directly. Clinically, the aspect ratios of cells inside cytopathological images provide important information for doctors to diagnose cancer. Therefore, it is difficult to resize directly. However, many existing studies have resized the images directly and have obtained highly robust classification results. To determine a reasonable interpretation, we have conducted a series of comparative experiments. First, the raw data of the SIPaKMeD dataset are pre-processed to obtain standard and scaled datasets. Then, the datasets are resized to 224 × 224 pixels. Finally, 22 deep learning models are used to classify the standard and scaled datasets. The results of the study indicate that deep learning models are robust to changes in the aspect ratio of cells in cervical cytopathological images. This conclusion is also validated via the Herlev dataset.