Search

Data augmentation is a crucial regularization technique for deep neural networks, particularly in medical imaging tasks with limited data. Deep learning models are highly effective at linearizing features, enabling the alteration of feature semantics through the shifting of latent space representations—an approach known as semantic data augmentation (SDA). The paradigm of SDA involves shifting features in a specified direction. Current SDA methods typically sample the amount of shifting from a Gaussian distribution or the sample variance. However, excessive shifting can lead to changes in data labels, which may negatively impact model performance. To address this issue, we propose a computationally efficient method called Bayesian Random Semantic Data Augmentation (BSDA). BSDA can be seamlessly integrated as a plug-and-play component into any neural network. Our experiments demonstrate that BSDA outperforms competitive methods and is suitable for both 2D and 3D medical image datasets, as well as most medical imaging modalities. Additionally, BSDA is compatible with mainstream neural network models and enhances baseline performance. The code is available online. [ABSTRACT FROM AUTHOR]

Currently, various deep learning methods have been developed to address the image enhancement tasks based on paired high-quality images as references. For the low-light endoscopic image enhancement task, it is difficult to obtain paired high-quality images and to extract features from dark areas. In addition, the enhanced images easily appear color distortions. In this study, we propose an unsupervised deep learning scheme based on the Cycle Generative Adversarial Network to enhance the endoscopic image. Because extracting features in the dark areas is important but challenging, we embedded an adaptive reverse attention module in generators to help the network focus on low-light areas and enhance these areas. We also introduce a color consistency constraint to maintain color constancy. To evaluate the performance of the proposed enhancement method, a blind evaluation methodology is proposed in view of no specific quality assessment metric specially designed on this field. Extensive subjective and objective experiment results demonstrate that the proposed method is competent for the colorectal endoscopic image enhancement task, and performs better than both conventional methods and popular deep learning-based methods on 200 real-captured colonoscopy images. In the objective experiment, the proposed method ranks first with a PIQE score of 11.1525 and an NIQE score of 11.1525, outperforming five competing methods. It also receives the best results from an average score of 1.455 over 200 test images of the subjective experiment. [ABSTRACT FROM AUTHOR]

Brain tumors are serious conditions, which can cause great trauma to patients, endangering their health and even leading to disability or death. Therefore, accurate preoperative diagnosis is particularly important. Accurate brain tumor segmentation based on deep learning plays an important role in the preoperative treatment planning process and has achieved good performance. However, one of the challenges involved is an insufficient ability to extract features with a large receptive field in encoder layers and guide the selection of deep semantic information in decoder layers. We propose small kernel convolution with an effective receptive field network (SCENet) based on UNet, which involves a small kernel convolution with effective receptive field shuffle module (SCER) and a channel spatial attention module (CSAM). The SCER module utilizes the inherent properties of stacking convolution to obtain effectively receptive fields and improve the features with a large receptive field extraction ability. CSAM of decoder layers can preserve more detailed features to capture clearer contours of the segmented image by calculating the weights of channels and spaces. An ASPP module is introduced to the bottleneck layer to enlarge the receptive field and can capture multi-scale detailed features. Furthermore, a large number of experiments were performed to evaluate the performance of our model on the BraTS2021 dataset. The SCENet achieved dice coefficient scores of 91.67%, 87.70%, and 83.35% for whole tumor (WT), tumor core (TC), and enhancing tumor (ET), respectively. The results show that the proposed model achieves the state-of-the-art performance compared with more than twelve benchmarks. [ABSTRACT FROM AUTHOR]

In the growing field of the Internet of Medical Things (IoMT), securing the transmission of medical images over public networks is a critical challenge. Medical images, being highly sensitive and often containing personally identifiable information, require robust protection against unauthorized access and tampering. This paper addresses this challenge by introducing a novel cryptosystem specifically tailored to the resource limitations inherent in IoMT environments. To meet the demand for protecting sensitive information in medical images, the proposed system integrates two layers of security: spatial watermarking and asymmetric encryption. At the core of our approach lies a newly developed, cost-effective spatial watermarking algorithm that seamlessly embeds watermarks within host images to facilitate tamper detection. Complementing this, we employ a resource-efficient Twin Message Fusion (TMF) encryption scheme to ensure confidentiality and integrity during transmission. The proposed cryptosystem is evaluated for both watermarking effectiveness and encryption robustness. Additionally, we assess the system's suitability for IoMT environments in terms of cost-effectiveness. Mean Squared Error (MSE) and Peak Signal-to-Noise Ratio (PSNR) metrics have been used to assess watermarking resilience against various attacks, while encryption efficacy is analyzed through histogram and entropy evaluations. Furthermore, we compare the complexity of our cryptosystem with five recently proposed techniques in both traditional and IoMT environments, demonstrating superior execution time, reduced image size, and enhanced encryption efficiency. [ABSTRACT FROM AUTHOR]

This paper explores the application of Visual Question-Answering (VQA) technology, which combines computer vision and natural language processing (NLP), in the medical domain, specifically for analyzing radiology scans. VQA can facilitate medical decision-making and improve patient outcomes by accurately interpreting medical imaging, which requires specialized expertise and time. The paper proposes developing an advanced VQA system for medical datasets using the Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation (BLIP) architecture from Salesforce, leveraging deep learning and transfer learning techniques to handle the unique challenges of medical/radiology images. The paper discusses the underlying concepts, methodologies, and results of applying the BLIP architecture and fine-tuning approaches for VQA in the medical domain, highlighting their effectiveness in addressing the complexities of VQA tasks for radiology scans. Inspired by the BLIP architecture from Salesforce, we propose a novel multi-modal fusion approach for medical VQA and evaluating its promising potential. [ABSTRACT FROM AUTHOR]

To serve a convenient healthcare network, storing medical images and diagnosis records in the cloud is a straightforward solution. Encrypting the medical images before uploading them to the cloud is a trivial strategy to prevent data breaches. However, the availability of images is also reduced. Since all images look noisy, one cannot obtain the intended image by a quick viewing of the file folder. In this paper, we propose a total solution to these considerations. A reversible data hiding scheme based on thumbnail-preserving image encryption is presented. By leveraging a multi-scale Haar wavelet transform and extended run-length coding, we can produce thumbnail-preserving encrypted medical images that provide a sufficient vacating room for embedding electronic patient records. The experimental results demonstrate that our scheme effectively balances the confidentiality and availability of encrypted medical images. It provides over 1 bit per pixel of capacity for embedding electronic patient records, making it suitable for cloud-based medical data management. [ABSTRACT FROM AUTHOR]

Automated segmentation of liver tumors on CT scans is essential for aiding diagnosis and assessing treatment. Computer-aided diagnosis can reduce the costs and errors associated with manual processes and ensure the provision of accurate and reliable clinical assessments. However, liver tumors in CT images vary significantly in size and have fuzzy boundaries, making it difficult for existing methods to achieve accurate segmentation. Therefore, this paper proposes MAEG-Net, a multi-scale adaptive feature fusion liver tumor segmentation network based on edge guidance. Specifically, we design a multi-scale adaptive feature fusion module that effectively incorporates multi-scale information to better guide the segmentation of tumors of different sizes. Additionally, to address the problem of blurred tumor boundaries in images, we introduce an edge-aware guidance module to improve the model's feature learning ability under these conditions. Evaluation results on the liver tumor dataset (LiTS2017) show that our method achieves a Dice coefficient of 71.84% and a VOE of 38.64%, demonstrating the best performance for liver tumor segmentation in CT images. [ABSTRACT FROM AUTHOR]

Image segmentation is a crucial task in the field of computer vision. Markov random fields (MRF) based image segmentation method can effectively capture intricate relationships among pixels. However, MRF typically requires an initial labeling field, and the number of classifications needs to be manually selected. To tackle these issues, we propose a novel medical image segmentation algorithm based on density peak clustering (DPC) and Markov random fields. Firstly, we improve DPC to make it applicable to grayscale images, named GIDPC. In the GIDPC method, local gray density and gray bias are defined to enable the automatic determination of the number of classifications. Then, GIDPC and MRF are combined to achieve image segmentation. Furthermore, a segmentation fusion method is employed to enhance the accuracy of image segmentation. We conduct comparison experiments on the whole brain atlas image library. Our proposed algorithm achieves high average values in uniformity measure, accuracy, precision and sensitivity, respectively. Experimental results demonstrate that the proposed algorithm outperforms other image segmentation methods. [ABSTRACT FROM AUTHOR]

Medical image denoising is essential for improving the clarity and accuracy of diagnostic images. In this paper, we present an enhanced wavelet-based method for medical image denoising, aiming to effectively remove noise while preserving critical image details. After applying wavelet denoising, a bilateral filter is utilized as a post-processing step to further enhance image quality by reducing noise while maintaining edge sharpness. The bilateral filter's effectiveness heavily depends on its parameters, which must be carefully optimized. To achieve this, we employ Bayesian optimization, a powerful technique that efficiently identifies the optimal filter parameters, ensuring the best balance between noise reduction and detail preservation. The experimental results demonstrate a significant improvement in image denoising performance, validating the effectiveness of our approach. [ABSTRACT FROM AUTHOR]

Background: Multi‐organ segmentation is a critical task in medical imaging, with wide‐ranging applications in both clinical practice and research. Accurate delineation of organs from high‐resolution 3D medical images, such as CT scans, is essential for radiation therapy planning, enhancing treatment outcomes, and minimizing radiation toxicity risks. Additionally, it plays a pivotal role in quantitative image analysis, supporting various medical research studies. Despite its significance, manual segmentation of multiple organs from 3D images is labor‐intensive and prone to low reproducibility due to high interoperator variability. Recent advancements in deep learning have led to several automated segmentation methods, yet many rely heavily on labeled data and human anatomy expertise. Purpose: In this study, our primary objective is to address the limitations of existing semi‐supervised learning (SSL) methods for abdominal multi‐organ segmentation. We aim to introduce a novel SSL approach that leverages unlabeled data to enhance the performance of deep neural networks in segmenting abdominal organs. Specifically, we propose a method that incorporates a redrawing network into the segmentation process to correct errors and improve accuracy. Methods: Our proposed method comprises three interconnected neural networks: a segmentation network for image segmentation, a teacher network for consistency regularization, and a redrawing network for object redrawing. During training, the segmentation network undergoes two rounds of optimization: basic training and readjustment. We adopt the Mean‐Teacher model as our baseline SSL approach, utilizing labeled and unlabeled data. However, recognizing significant errors in abdominal multi‐organ segmentation using this method alone, we introduce the redrawing network to generate redrawn images based on CT scans, preserving original anatomical information. Our approach is grounded in the generative process hypothesis, encompassing segmentation, drawing, and assembling stages. Correct segmentation is crucial for generating accurate images. In the basic training phase, the segmentation network is trained using both labeled and unlabeled data, incorporating consistency learning to ensure consistent predictions before and after perturbations. The readjustment phase focuses on reducing segmentation errors by optimizing the segmentation network parameters based on the differences between redrawn and original CT images. Results: We evaluated our method using two publicly available datasets: the beyond the cranial vault (BTCV) segmentation dataset (training: 44, validation: 6) and the abdominal multi‐organ segmentation (AMOS) challenge 2022 dataset (training:138, validation:16). Our results were compared with state‐of‐the‐art SSL methods, including MT and dual‐task consistency (DTC), using the Dice similarity coefficient (DSC) as an accuracy metric. On both datasets, our proposed SSL method consistently outperformed other methods, including supervised learning, achieving superior segmentation performance for various abdominal organs. These findings demonstrate the effectiveness of our approach, even with a limited number of labeled data. Conclusions: Our novel semi‐supervised learning approach for abdominal multi‐organ segmentation addresses the challenges associated with this task. By integrating a redrawing network and leveraging unlabeled data, we achieve remarkable improvements in accuracy. Our method demonstrates superior performance compared to existing SSL and supervised learning methods. This approach holds great promise in enhancing the precision and efficiency of multi‐organ segmentation in medical imaging applications. [ABSTRACT FROM AUTHOR]

Over 8% of women will be diagnosed with breast tumors (BT) in their lifetime. Tumors are formed by the uncontrollable development of tissues in a specific area of the body. They can be benign or malignant. The best survival rate can be expected with earlier screening and diagnosis of the tumor. To distinguish between benign and malignant tumors in x-ray images of the breast, segmentation of the tumor is a crucial first step. Screening mammography is an efficient method of detecting BT. As a result, the research presented two distinct deep learning models, termed SegNet and UNet architectures, to segment BT from mammograms. Datasets accessible to the public were utilized in the proposed system, specifically INbreast. Histogram equalization is used on datasets during preprocessing to improve the compressed areas and normalize the pixel dispersion. To avoid overfitting and boost the quantity of training data, augmentation techniques are employed. The metrics like the dice coefficient (DC) and the Intersection of Union (IoU) score are considered to evaluate the model. The metrics of the SegNets model are greater than the U-Net, as demonstrated by the experimental results. For the INbreast dataset, the SegNets achieve a maximum DC of 92.75% and an IoU score of 86.49%. [ABSTRACT FROM AUTHOR]

Active learning (AL) algorithms are increasingly being used to train models with limited data for annotation tasks. However, the selection of data for AL is a complex issue due to the restricted information on unseen data. To tackle this problem, a technique we refer to as Partial Image Active Annotation (PIAA) employs the edge information of unseen images as prior knowledge to gauge uncertainty. This uncertainty is determined by examining the divergence and entropy in model predictions across edges. The resulting measure is then applied to choose superpixels from input images for active annotation. We demonstrate the effectiveness of PIAA in multi-class Optical Coherence Tomography (OCT) segmentation tasks, attaining a Dice score comparable to state-of-the-art OCT segmentation algorithms trained with extensive annotated data. Concurrently, we successfully reduce annotation label costs to 12%, 2.3%, and 3%, respectively, across three publicly accessible datasets (Duke, AROI, and UMN). [ABSTRACT FROM AUTHOR]

Manual annotation of medical image datasets is labor-intensive and prone to biases. Moreover, the rate at which image data accumulates significantly outpaces the speed of manual annotation, posing a challenge to the advancement of machine learning, particularly in the realm of supervised learning. Self-supervised learning is an emerging field that capitalizes on unlabeled data for training, thereby circumventing the need for extensive manual labeling. This learning paradigm generates synthetic pseudo-labels through pretext tasks, compelling the network to acquire image representations in a pseudo-supervised manner and subsequently fine-tuning with a limited set of annotated data to achieve enhanced performance. This review begins with an overview of prevalent types and advancements in self-supervised learning, followed by an exhaustive and systematic examination of methodologies within the medical imaging domain from 2018 to September 2024. The review encompasses a range of medical image modalities, including CT, MRI, X-ray, Histology, and Ultrasound. It addresses specific tasks, such as Classification, Localization, Segmentation, Reduction of False Positives, Improvement of Model Performance, and Enhancement of Image Quality. The analysis reveals a descending order in the volume of related studies, with CT and MRI leading the list, followed by X-ray, Histology, and Ultrasound. Except for CT and MRI, there is a greater prevalence of studies focusing on contrastive learning methods over generative learning approaches. The performance of MRI/Ultrasound classification and all image types segmentation still has room for further exploration. Generally, this review can provide conceptual guidance for medical professionals to combine self-supervised learning with their research. [ABSTRACT FROM AUTHOR]

Deep learning-based segmentation models often struggle to achieve optimal performance when encountering new, unseen semantic classes. Their effectiveness hinges on vast amounts of annotated data and high computational resources for training. However, a promising solution to mitigate these challenges is the adoption of few-shot segmentation (FSS) networks, which can train models with reduced annotated data. The inherent complexity of medical images limits the applicability of FSS in medical imaging, despite its potential. Recent advancements in self-supervised label-efficient FSS models have demonstrated remarkable efficacy in medical image segmentation tasks. This paper presents a novel FSS architecture that enhances segmentation accuracy by utilising fewer features than existing methodologies. Additionally, this paper proposes a novel self-supervised learning approach that utilises supervoxel and augmented superpixel images to further enhance segmentation accuracy. This paper assesses the efficacy of the proposed model on two different datasets: abdominal magnetic resonance imaging (MRI) and cardiac MRI. The proposed model achieves a mean dice score and mean intersection over union of 81.62% and 70.38% for abdominal images, and 79.38% and 65.23% for cardiac images. [ABSTRACT FROM AUTHOR]

The use of artificial intelligence (AI) in the segmentation of liver structures in medical images has become a popular research focus in the past half-decade. The performance of AI tools in screening for this task may vary widely and has been tested in the literature in various datasets. However, no scientometric report has provided a systematic overview of this scientific area. This article presents a systematic and bibliometric review of recent advances in neuronal network modeling approaches, mainly of deep learning, to outline the multiple research directions of the field in terms of algorithmic features. Therefore, a detailed systematic review of the most relevant publications addressing fully automatic semantic segmenting liver structures in Computed Tomography (CT) images in terms of algorithm modeling objective, performance benchmark, and model complexity is provided. The review suggests that fully automatic hybrid 2D and 3D networks are the top performers in the semantic segmentation of the liver. In the case of liver tumor and vasculature segmentation, fully automatic generative approaches perform best. However, the reported performance benchmark indicates that there is still much to be improved in segmenting such small structures in high-resolution abdominal CT scans. [ABSTRACT FROM AUTHOR]

Moyamoya disease (MMD) is a complex cerebrovascular disease with unknown etiology. At present, it is mainly diagnosed by imaging examination. With the continuous development of imaging technology and the emergence of artificial intelligence, it has made great contributions to the diagnosis, identification and risk factor analysis of MMD. This paper focuses on the specific application of machine learning and deep learning algorithms in the field of artificial intelligence in the imaging of MMD. This paper summarizes and analyzes the shortcomings of such research at present and the prospects for the future. [ABSTRACT FROM AUTHOR]

Magnetic resonance imaging (MRI) creates detailed images by combining a powerful magnetic field with high-frequency radio waves. MRI images need to increase brightness and contrast to facilitate distinguishing diseases. This study aims to improve the MRI images depending on the suggested algorithm, including three main stages. The first is to divide the image into three main areas using Multilevel image thresholds and then improve using CLAHE for each area separately. Finally, these images combine to form one image and improve it again. By analyzing the results and calculating non-referenced quality measures, the proposed method obtained the best quality results compared to the rest of the methods with quality rates of EN(6.14), AG(6.79), and CEM(0.84), which indicates excellent success in increasing clarity in those images. [ABSTRACT FROM AUTHOR]

As deep learning technology has progressed, automated medical image analysis is becoming ever more crucial in clinical diagnosis. However, due to the diversity and complexity of blood cell images, traditional models still exhibit deficiencies in blood cell detection. To address blood cell detection, we developed the TW-YOLO approach, leveraging multi-scale feature fusion techniques. Firstly, traditional CNN (Convolutional Neural Network) convolution has poor recognition capabilities for certain blood cell features, so the RFAConv (Receptive Field Attention Convolution) module was incorporated into the backbone of the model to enhance its capacity to extract geometric characteristics from blood cells. At the same time, utilizing the feature pyramid architecture of YOLO (You Only Look Once), we enhanced the fusion of features at different scales by incorporating the CBAM (Convolutional Block Attention Module) in the detection head and the EMA (Efficient Multi-Scale Attention) module in the neck, thereby improving the recognition ability of blood cells. Additionally, to meet the specific needs of blood cell detection, we designed the PGI-Ghost (Programmable Gradient Information-Ghost) strategy to finely describe the gradient flow throughout the process of extracting features, further improving the model's effectiveness. Experiments on blood cell detection datasets such as BloodCell-Detection-Dataset (BCD) reveal that TW-YOLO outperforms other models by 2%, demonstrating excellent performance in the task of blood cell detection. In addition to advancing blood cell image analysis research, this work offers strong technical support for future automated medical diagnostics. [ABSTRACT FROM AUTHOR]

Abstract Automated segmentation of liver tumors on CT scans is essential for aiding diagnosis and assessing treatment. Computer-aided diagnosis can reduce the costs and errors associated with manual processes and ensure the provision of accurate and reliable clinical assessments. However, liver tumors in CT images vary significantly in size and have fuzzy boundaries, making it difficult for existing methods to achieve accurate segmentation. Therefore, this paper proposes MAEG-Net, a multi-scale adaptive feature fusion liver tumor segmentation network based on edge guidance. Specifically, we design a multi-scale adaptive feature fusion module that effectively incorporates multi-scale information to better guide the segmentation of tumors of different sizes. Additionally, to address the problem of blurred tumor boundaries in images, we introduce an edge-aware guidance module to improve the model's feature learning ability under these conditions. Evaluation results on the liver tumor dataset (LiTS2017) show that our method achieves a Dice coefficient of 71.84% and a VOE of 38.64%, demonstrating the best performance for liver tumor segmentation in CT images.

Abstract Manual annotation of medical image datasets is labor-intensive and prone to biases. Moreover, the rate at which image data accumulates significantly outpaces the speed of manual annotation, posing a challenge to the advancement of machine learning, particularly in the realm of supervised learning. Self-supervised learning is an emerging field that capitalizes on unlabeled data for training, thereby circumventing the need for extensive manual labeling. This learning paradigm generates synthetic pseudo-labels through pretext tasks, compelling the network to acquire image representations in a pseudo-supervised manner and subsequently fine-tuning with a limited set of annotated data to achieve enhanced performance. This review begins with an overview of prevalent types and advancements in self-supervised learning, followed by an exhaustive and systematic examination of methodologies within the medical imaging domain from 2018 to September 2024. The review encompasses a range of medical image modalities, including CT, MRI, X-ray, Histology, and Ultrasound. It addresses specific tasks, such as Classification, Localization, Segmentation, Reduction of False Positives, Improvement of Model Performance, and Enhancement of Image Quality. The analysis reveals a descending order in the volume of related studies, with CT and MRI leading the list, followed by X-ray, Histology, and Ultrasound. Except for CT and MRI, there is a greater prevalence of studies focusing on contrastive learning methods over generative learning approaches. The performance of MRI/Ultrasound classification and all image types segmentation still has room for further exploration. Generally, this review can provide conceptual guidance for medical professionals to combine self-supervised learning with their research.

With the development of computer vision and medical image processing technology, lung CT scan image segmentation plays an increasingly important role in clinical diagnosis. Doctors can receive precise anatomical structure information from the accurate segmentation of complicated lung structures from CT scan images, which can help with disease diagnosis and treatment planning. However, the complexity and diversity of lung structures make it difficult to design and optimize segmentation algorithms, especially in the presence of lesions, masses, or scars. To solve this problem, the study proposes an enhanced dung beetle optimizer with the fractional order calculus strategy (FDBO) to identify the lung CT scan image’s optimal thresholds. In the FDBO, the fractional order calculus strategy is used to store past individual information and apply it to future individuals to obtain higher-quality individuals. Additionally, adaptive technology is introduced to balance the algorithm’s ability between exploration and exploitation, and the global best position is mutated to enhance the ability of the algorithm to escape from the local optimum. To evaluate the performance of the FDBO, on the one hand, the FDBO is used to solve the CEC2019 benchmark functions in the function global optimization task. On the other hand, the FDBO is applied to the multilevel threshold image segmentation problem with the Otsu method, and the inter-class variance is used as the objective function to solve for its maximum. The experiment’s numerical results show that the FDBO has excellent optimization accuracy and stability. When the threshold is high, the FDBO can obtain better segmentation quality than the other seven algorithms for most lung CT scan images. [ABSTRACT FROM AUTHOR]

Abstract This paper introduces the efficient medical-images-aimed segment anything model (EMedSAM), addressing the high computational demands and limited adaptability of using SAM for medical image segmentation tasks. We present a novel, compact image encoder, DD-TinyViT, designed to enhance segmentation efficiency through an innovative parameter tuning method called med-adapter. The lightweight DD-TinyViT encoder is derived from the well-known ViT-H using a decoupled distillation approach.The segmentation and recognition capabilities of EMedSAM for specific structures are improved by med-adapter, which dynamically adjusts the model parameters specifically for medical imaging. We conducted extensive testing on EMedSAM using the public FLARE 2022 dataset and datasets from the First Hospital of Zhejiang University School of Medicine. The results demonstrate that our model outperforms existing state-of-the-art models in both multi-organ and lung segmentation tasks.

Abstract Traditionally, constructing training datasets for automatic muscle segmentation from medical images involved skilled operators, leading to high labor costs and limited scalability. To address this issue, we developed a tool that enables efficient annotation by non-experts and assessed its effectiveness for training an automatic segmentation network. Our system allows users to deform a template three-dimensional (3D) anatomical model to fit a target magnetic-resonance image using free-form deformation with independent control points for axial, sagittal, and coronal directions. This method simplifies the annotation process by allowing non-experts to intuitively adjust the model, enabling simultaneous annotation of all muscles in the template. We evaluated the quality of the tool-assisted segmentation performed by non-experts, which achieved a Dice coefficient greater than 0.75 compared to expert segmentation, without significant errors such as mislabeling adjacent muscles or omitting musculature. An automatic segmentation network trained with datasets created using this tool demonstrated performance comparable to or superior to that of networks trained with expert-generated datasets. This innovative tool significantly reduces the time and labor costs associated with dataset creation for automatic muscle segmentation, potentially revolutionizing medical image annotation and accelerating the development of deep learning-based segmentation networks in various clinical applications.

Several medical consultations and examinations have been undertaken online since the Covid outbreak. However, when private data was communicated over the internet or uploaded to the cloud, medical information became more susceptible to security risks. Steganography is a technique that can be used to hide sensitive information within a cover image. This paper presents four improved algorithms to enhance steganography's performance in medical images. A full-size hidden image that is as huge as a cover image cannot be handled by previous methods, which is what the algorithmic design is meant to address. Several creative methods are presented, including the computation of Discrete Cosine Transform (DCT) coefficients based on scaled floating values, the addition of an adaptive compression matrix, and a new approach for systematically dispersing a concealed number of bits across multiple separate locations in the cover image. The results of the experiment showed a notable advancement over the earlier research. Our secret image size is substantially larger than the past studies, yet the structure similarity index matrix (SSIM) of the best reconstructed secret image is close to ideal, the peak signal-to-noise ratio (PSNR) and the payload capacity are higher than in the previous studies. This research is beneficial since it contributes to a medical application for enhancing the security of information concealed in chest X-ray images. Medical personnel can generate an image that conceals patient information in a secure manner. [ABSTRACT FROM AUTHOR]

The evaluation of carotid ultrasound image plaque requires a large number of experienced clinicians, and the ultrasound image has the characteristics of blurred boundary and strong noise interference, making the evaluation of plaques time-consuming and laborious. Therefore, a fully automated carotid plaque segmentation method is urgently needed to solve the problem of manpower scarcity. This study proposes a deep neural network model based on Swin-Transformer (Shifted-Windows Transformer) block for the automatic segmentation of carotid plaques. Based on the U-Net(U-Convolutional Network) architecture, the encoding part uses three convolutional blocks for image down-sampling to obtain feature images of different resolution sizes, and then adds six pairs of two consecutive Swin-Transformer blocks for more refined feature extraction. The decoding part up-samples the refined features output by the Swin-Transformer module step by step, and jump-joints them with the feature maps of each resolution level in the encoding part, respectively. The comparison experiments based on the data set of Tong Ren Hospital show that the Dice index of the proposed deep neural network model reaches 0.814 2, which is higher than that of other comparison networks. The results demonstrate that the proposed model can effectively extract the features of carotid ultrasound image plaques and achieve automated and high-precision plaque segmentation. [ABSTRACT FROM AUTHOR]

Gastrointestinal (GI) diseases are a major issue in the human digestive system. Therefore, many studies have explored the automatic classification of GI diseases to reduce the burden on clinicians and improve patient outcomes for both diagnosis and treatment purposes. Convolutional neural networks (CNNs) and Vision Transformers (ViTs) in deep learning approaches have become a popular research area for the automatic detection of diseases from medical images. This study evaluated the classification performance of thirteen different CNN models and two different ViT architectures on endoscopic images. The impact of transfer learning parameters on classification performance was also observed. The tests revealed that the classification accuracies of the ViT models were 91.25% and 90.50%, respectively. In contrast, the DenseNet201 architecture, with optimized transfer learning parameters, achieved an accuracy of 93.13%, recall of 93.17%, precision of 93.13%, and an F1 score of 93.11%, making it the most successful model among all the others. Considering the results, it is evident that a well-optimized CNN model achieved better classification performance than the ViT models. [ABSTRACT FROM AUTHOR]

How to guarantee the confidentiality of sensitive data communicated over the Internet and restrict access to designated information is today's key security and protection concern in telemedicine. In this work, we suggest a reliable and blind medical image watermarking method that combines integer wavelet transform (IWT) and singular value decomposition to keep such information private. A major drawback of current IWT-based watermarking systems is their low embedding capacity. This paper suggests an IWT-based secure large capacity watermarking solution to overcome this specific drawback. The proposed technique effectively preserves a considerable quality of watermarked images, and the watermark is resistant to the most frequently used attacks in watermarking, according to experiment results on imperceptibility and robustness. [ABSTRACT FROM AUTHOR]

Medical image encryption has currently drawn particular attention because of the sensitive nature of medical data, the lack of efficient image encryption employing innovative encryption methods. To enhance the encryption of medical images, a number of encryption approaches were suggested and created. This research proposes a hyperchaotic map as well as a deep learning (DL) based selective medical image encryption technique. Initially, the 3D_Att_ResU-Net segments more sensitive region of interest (ROI) which is the brain tumours. The process continues with a hyperchaotic map for encrypting the ROI of plain medical image. On the other hand, the region of non interest (RONI) is compressed based on discrete cosine transform (DCT). The encrypting of ROI as well as compression of the background could provide the best possible balance between encryption performance and security. The samples for the simulation included a variety of common medical image types as well as a few normal images. The outcomes have been analysed using standard image cryptanalysis techniques. The findings demonstrated that cipher-images had uniformly distributed histogram, high information entropy, low correlation between neighbouring pixels, and good visual quality. The method has a large key space as well as low time complexity, yet it is sensitive to the plain image and the initial key. Interestingly, it has been found that the values of NPCR and UACI are higher than 99.75% and 36. 53%, respectively besides large key space registered as an order of 2339. The entropy of the Region of Interest (ROI) is roughly estimated to be 8, which is significantly represent a better security for the current system. The estimated encryption time required for the ROI of size 8.09 KB is 0.3s and to compress the RONI of size 73.81 is 1.02s while the total time required to encrypt whole image of size 81.9 is 2.4737s making it suitable for real time applications. The results depict that the current presented encryption scheme gives high security with less computational time and complexity. [ABSTRACT FROM AUTHOR]

Content Based Image Retrieval (CBIR) is a relatively new idea in the field of real-time image retrieval applications; it is a framework for retrieving pictures from diverse medical imaging sources using a variety of image-related attributes, such as color, texture, and form. Using both single and multiple input queries, CBIR processes semantic data or the same object for various class labels in the context of medical image retrieval. Due to the ambiguity of image search, optimizing the retrieval of a query picture by comparing it across numerous image sources may be problematic. The goal is to find a way to optimize the process by which requested images are retrieved from various storage locations. To effectively extract medical images, we propose a hybrid framework (consisting of deep convolution neural networks (DCNN) and the Pareto Optimization technique). In order to obtain medical pictures, a DCNN is trained on them, and then its properties and classification results are employed. Explore enhanced effective medical picture retrieval by using a Pareto optimization strategy to eliminate superfluous and dominant characteristics. When it comes to retrieving images by query from various picture archives, our method outperforms more conventional methods. Use the jargon of machine learning to propose a Novel Unsupervised Label Indexing (NULI) strategy for retrieving picture labels. To enhance the effectiveness of picture retrieval, we characterize machine learning as a matrix convex optimization using a cluster rebased matrix representation. We describe an empirical investigation on many medical picture datasets, finding that the searchbased image annotation (SBIA) schema benefits from our suggested method. As a result, CT images of the lung region are explored in this study by constructing a content-based image retrieval system using various machine learning and Artificial Intelligence techniques. Real-world applications of medical imaging are becoming more significant. Medical research facilities acquire and archive a wide variety of medical pictures digitally. [ABSTRACT FROM AUTHOR]

The International Skin Imaging Collaboration (ISIC) datasets are pivotal resources for researchers in machine learning for medical image analysis, especially in skin cancer detection. These datasets contain tens of thousands of dermoscopic photographs, each accompanied by gold-standard lesion diagnosis metadata. Annual challenges associated with ISIC datasets have spurred significant advancements, with research papers reporting metrics surpassing those of human experts. Skin cancers are categorized into melanoma and non-melanoma types, with melanoma posing a greater threat due to its rapid potential for metastasis if left untreated. This paper aims to address challenges in skin cancer detection via visual inspection and manual examination of skin lesion images, processes historically known for their laboriousness. Despite notable advancements in machine learning and deep learning models, persistent challenges remain, largely due to the intricate nature of skin lesion images. We review research on convolutional neural networks (CNNs) in skin cancer classification and segmentation, identifying issues like data duplication and augmentation problems. We explore the efficacy of Vision Transformers (ViTs) in overcoming these challenges within ISIC dataset processing. ViTs leverage their capabilities to capture both global and local relationships within images, reducing data duplication and enhancing model generalization. Additionally, ViTs alleviate augmentation issues by effectively leveraging original data. Through a thorough examination of ViT-based methodologies, we illustrate their pivotal role in enhancing ISIC image classification and segmentation. This study offers valuable insights for researchers and practitioners looking to utilize ViTs for improved analysis of dermatological images. Furthermore, this paper emphasizes the crucial role of mathematical and computational modeling processes in advancing skin cancer detection methodologies, highlighting their significance in improving algorithmic performance and interpretability. [ABSTRACT FROM AUTHOR]

The prediction of response to breast cancer therapy involves assessing the effectiveness of treatment by comparing biomarker levels before and after treatment. Deep learning (DL) models can provide a non-invasive and early way to evaluate the response to therapy based on medical imaging analysis. We conducted this systematic review to investigate the current DL based methods for predicting breast cancer therapy response using medical imaging. This review included 19 studies based on the PRISMA methodology. Some selected studies personalized the Convolutional Neural Network (CNN) architecture to improve its performance in handling medical images, while others used pre-trained models. The accuracy rates range from 0.73 to 0.90, and the Area Under the Curve (AUC) reaches 0.98. Our study's findings suggest that the performance of these approaches varies depending on various medical imaging modalities, the nature of the DL architecture used, and the fusion of training data sources. However, several challenges related to their explainability and generalizability arise. Therefore, it is necessary to develop larger datasets and broaden the scope of current studies to include multi-center studies. [ABSTRACT FROM AUTHOR]

Recently, foundation models have been introduced demonstrating various tasks in the field of computer vision. These models such as Segment Anything Model (SAM) are generalized models trained using huge datasets. Currently, ongoing research focuses on exploring the effective utilization of these generalized models for Specific domains, such as medical imaging. However, in medical imaging, the lack of training samples due to privacy concerns and other factors presents a major challenge for applying these generalized models to medical image segmentation task. To address this issue, the effective fine tuning of these models is crucial to ensure their optimal utilization. In this study, we propose to combine a complementary Convolutional Neural Network (CNN) along with the standard SAM network for medical image segmentation. To reduce the burden of fine tuning large foundation model and implement cost-efficient training scheme, we focus only on fine-tuning the additional CNN network and SAM decoder part. This strategy significantly reduces training time and achieves competitive results on publicly available dataset. The code is available at ">https://github.com/11yxk/SAM-LST. [ABSTRACT FROM AUTHOR]

Deep learning has been the subject of a significant amount of research interest in the development of novel algorithms for deep learning algorithms and medical image processing have proven very effective in a number of medical imaging tasks to help illness identification and diagnosis. The shortage of large-sized datasets that are also adequately annotated is a key barrier that is preventing the continued advancement of deep learning models used in medical image analysis, despite the effectiveness of these models. Over the course of the previous 5 years, a great number of research have concentrated on finding solutions to this problem. In this work, we present a complete overview of the use of deep learning techniques in a variety of medical image analysis tasks by reviewing and summarizing the current research that have been conducted in this area. In particular, we place an emphasis on the most recent developments and contributions of state-of-the-art semi-supervised and unsupervised deep learning in medical image analysis. These advancements and contributions are shortened based on various application scenarios, which include image registration, segmentation, classification and detection. In addition to this, we explore the significant technological obstacles that lie ahead and provide some potential answers for the ongoing study. [ABSTRACT FROM AUTHOR]

This systematic review aims to assess the efficacy of Artificial Intelligence (AI) applications in orthopedic surgery, with a focus on diagnostic accuracy and outcome prediction. In this review, we expose the findings of a systematic literature review awning the papers published from 2016 to October 2023 where authors worked on the application of an AI techniques and methods to an orthopedic purpose or problem. After application of inclusion and exclusion criteria on the extracted papers from PubMed and Google Scholar databases, 75 studies were included in this review. We examined, screened, and analyzed their content according to PRISMA guidelines. We also extracted data about the study design, the datasets included in the experiment, the reported performance measures and the results obtained. In this report, we will share the results of our survey by outlining the key machine and Deep Learning (DL) techniques, such as Convolutional Neural Network (CNN), Autoencoders and Generative Adversarial Network, that were mentioned, the various application domains in orthopedics, the type of source data and its modality, as well as the overall quality of their predictive capabilities. We aim to describe the content of the articles in detail and provide insights into the most notable trends and patterns observed in the survey data. [ABSTRACT FROM AUTHOR]

Introduction: Soft tissue sarcomas, similar in incidence to cervical and esophageal cancers, arise from various soft tissues like smooth muscle, fat, and fibrous tissue. Effective segmentation of sarcomas in imaging is crucial for accurate diagnosis. Methods: This study collected multi-modal MRI images from 45 patients with thigh soft tissue sarcoma, totaling 8,640 images. These images were annotated by clinicians to delineate the sarcoma regions, creating a comprehensive dataset. We developed a novel segmentation model based on the UNet framework, enhanced with residual networks and attention mechanisms for improved modality-specific information extraction. Additionally, self-supervised learning strategies were employed to optimize feature extraction capabilities of the encoders. Results: The new model demonstrated superior segmentation performance when using multi-modal MRI images compared to single-modal inputs. The effectiveness of the model in utilizing the created dataset was validated through various experimental setups, confirming the enhanced ability to characterize tumor regions across different modalities. Discussion: The integration of multi-modal MRI images and advanced machine learning techniques in our model significantly improves the segmentation of soft tissue sarcomas in thigh imaging. This advancement aids clinicians in better diagnosing and understanding the patient's condition, leveraging the strengths of different imaging modalities. Further studies could explore the application of these techniques to other types of soft tissue sarcomas and additional anatomical sites. [ABSTRACT FROM AUTHOR]

Traditionally, constructing training datasets for automatic muscle segmentation from medical images involved skilled operators, leading to high labor costs and limited scalability. To address this issue, we developed a tool that enables efficient annotation by non-experts and assessed its effectiveness for training an automatic segmentation network. Our system allows users to deform a template three-dimensional (3D) anatomical model to fit a target magnetic-resonance image using free-form deformation with independent control points for axial, sagittal, and coronal directions. This method simplifies the annotation process by allowing non-experts to intuitively adjust the model, enabling simultaneous annotation of all muscles in the template. We evaluated the quality of the tool-assisted segmentation performed by non-experts, which achieved a Dice coefficient greater than 0.75 compared to expert segmentation, without significant errors such as mislabeling adjacent muscles or omitting musculature. An automatic segmentation network trained with datasets created using this tool demonstrated performance comparable to or superior to that of networks trained with expert-generated datasets. This innovative tool significantly reduces the time and labor costs associated with dataset creation for automatic muscle segmentation, potentially revolutionizing medical image annotation and accelerating the development of deep learning-based segmentation networks in various clinical applications. [ABSTRACT FROM AUTHOR]

The classification of medical images enables physicians to perform expeditious and accurate data analysis, increasing the chances of timely disease diagnosis and early intervention to the patient. However, classification is a time‐consuming and labour intensive process when done manually. The Capsule Network (CapsNet) architecture has advantages in accurately and quickly classifying medical images due to its ability to evaluate images within part‐whole relationships, robustness to data rotations and affine transformations, and good performance on small datasets. However, CapsNet may demonstrate low performance on complex datasets. In this study, a new CapsNet model named MResCaps is proposed to overcome this disadvantage and enhance its performance on complex images. MResCaps utilizes an increasing number of residual blocks in each layer in parallel lane to obtain rich feature maps at different levels, aiming to achieve high success in the classification of various medical images. To evaluate the model's performance, the CIFAR10 dataset and the DermaMNIST, PneumoniaMNIST, and OrganMNIST‐S datasets from the MedMNIST dataset collection are used. MResCaps outperformed CapsNet by 20% in terms of accuracy on the CIFAR10 dataset. In addition, AUC values of 96.25%, 96.30%, and 97.12% were achieved in DermaMNIST, PneumoniaMNIST, and OrganMNIST‐S datasets, respectively. The results show that the proposed new model MResCaps improves the performance of CapsNet in the classification of complex and medical images. Furthermore, the model has demonstrated a better performance in comparison with extant studies in the literature. This study aims to contribute significantly to the literature by introducing a novel perspective on CapsNet‐based architectures for the classification of medical images through a parallel‐laned architecture and a rich feature capsule‐focused approach. [ABSTRACT FROM AUTHOR]

The application of magnetic resonance imaging (MRI) in the classification of brain tumors is constrained by the complex and time-consuming characteristics of traditional diagnostics procedures, mainly because of the need for a thorough assessment across several regions. Nevertheless, advancements in deep learning (DL) have facilitated the development of an automated system that improves the identification and assessment of medical images, effectively addressing these difficulties. Convolutional neural networks (CNNs) have emerged as steadfast tools for image classification and visual perception. This study introduces an innovative approach that combines CNNs with a hybrid attention mechanism to classify primary brain tumors, including glioma, meningioma, pituitary, and no-tumor cases. The proposed algorithm was rigorously tested with benchmark data from well-documented sources in the literature. It was evaluated alongside established pre-trained models such as Xception, ResNet50V2, Densenet201, ResNet101V2, and DenseNet169. The performance metrics of the proposed method were remarkable, demonstrating classification accuracy of 98.33%, precision and recall of 98.30%, and F1-score of 98.20%. The experimental finding highlights the superior performance of the new approach in identifying the most frequent types of brain tumors. Furthermore, the method shows excellent generalization capabilities, making it an invaluable tool for healthcare in diagnosing brain conditions accurately and efficiently. [ABSTRACT FROM AUTHOR]

In the field of dentistry, the presence of dental calculus is a commonly encountered issue. If not addressed promptly, it has the potential to lead to gum inflammation and eventual tooth loss. Bitewing (BW) images play a crucial role by providing a comprehensive visual representation of the tooth structure, allowing dentists to examine hard-to-reach areas with precision during clinical assessments. This visual aid significantly aids in the early detection of calculus, facilitating timely interventions and improving overall outcomes for patients. This study introduces a system designed for the detection of dental calculus in BW images, leveraging the power of YOLOv8 to identify individual teeth accurately. This system boasts an impressive precision rate of 97.48%, a recall (sensitivity) of 96.81%, and a specificity rate of 98.25%. Furthermore, this study introduces a novel approach to enhancing interdental edges through an advanced image-enhancement algorithm. This algorithm combines the use of a median filter and bilateral filter to refine the accuracy of convolutional neural networks in classifying dental calculus. Before image enhancement, the accuracy achieved using GoogLeNet stands at 75.00%, which significantly improves to 96.11% post-enhancement. These results hold the potential for streamlining dental consultations, enhancing the overall efficiency of dental services. [ABSTRACT FROM AUTHOR]

This work evaluates GPT-4V's multimodal capability for medical image analysis, focusing on three representative tasks radiology report generation, medical visual question answering, and medical visual grounding. For the evaluation, a set of prompts is designed for each task to induce the corresponding capability of GPT-4V to produce sufficiently good outputs. Three evaluation ways including quantitative analysis, human evaluation, and case study are employed to achieve an in-depth and extensive evaluation. Our evaluation shows that GPT-4V excels in understanding medical images can generate high-quality radiology reports and effectively answer questions about medical images. Meanwhile, it is found that its performance for medical visual grounding needs to be substantially improved. In addition, we observe the discrepancy between the evaluation outcome from quantitative analysis and that from human evaluation. This discrepancy suggests the limitations of conventional metrics in assessing the performance of large language models like GPT-4V and the necessity of developing new metrics for automatic quantitative analysis.

Memristors are considered one of the most promising new-generation memory technologies due to their high integration density, fast read/write speeds, and ultra-low power consumption. Natural biomaterials have attracted interest in integrated circuits and electronics because of their environmental friendliness, sustainability, low cost, and excellent biocompatibility. In this study, a sustainable biomemristor with Ag/mugwort:PVDF/ITO structure was prepared using spin-coating and magnetron sputtering methods, which exhibited excellent durability, significant resistance switching (RS) behavior and unidirectional conduction properties when three metals were used as top electrode. By studying the conductivity mechanism of the device, a charge conduction model was established by the combination of F-N tunneling, redox, and complexation reaction. Finally, the novel logic gate circuits were constructed using the as-prepared memristor, and further memristor based encryption circuit using 3-8 decoder was innovatively designed, which can realize uniform rule encryption and decryption of medical information for data and medical images. Therefore, this work realizes the integration of memristor with traditional electronic technology and expands the applications of sustainable biomemristors in digital circuits, data encryption, and medical image security.