Search

Medical image segmentation is very essential for computer-aided diagnosis in the field of medical imaging. In the last decade, Deep Learning-based frameworks (e.g., UNet) have been widely used in medical applications such as image segmentation tasks. Recently, numerous Transformer-based frameworks are presented for the image segmentation tasks as their design can utilize long-range dependencies. Transformer's design has a weak inductive bias since it does not take advantage of local relationships between pixels and lacks scale invariance. Consequently, Transformers require large datasets for convergence whereas the availability of massive medical datasets is challenging. In this paper, we present a graph-based approach replacing Transformer design to capture long-range dependencies and reduce computational cost. Our proposed framework achieves competitive performance using publicly available dataset Synapse.

Developing a fast and accurate classifier is an important part of a computer-aided diagnosis system for skin cancer. Melanoma is the most dangerous form of skin cancer which has a high mortality rate. Early detection and prognosis of melanoma can improve survival rates. In this paper, we propose a deep convolutional neural network for automated melanoma detection that is scalable to accommodate a variety of hardware and software constraints. Dermoscopic skin images collected from open sources were used for training the network. The trained network was then tested on a dataset of 2150 malignant or benign images. Overall, the classifier achieved high average values for accuracy, sensitivity, and specificity of 82.95%, 82.99%, and 83.89% respectively. It outperfomed other exisitng networks using the same dataset.

Major Depressive Disorder (MDD) is a very serious mental illness that can affect the daily lives of patients. Accurate diagnosis of this disorder is necessary for planning individualized treatment. However, diagnosing MDD requires the clinicians to personally interview the subjects and rate the symptoms based on Diagnostic and Statistical Manual of Mental Disorders (DSM), which can be very time consuming. Discovering quantifiable signals and biomarkers associated with MDD using functional magnetic resonance imaging (fMRI) scans of patients have the potential to assist the clinicians in their assessment. This paper explores the use of resting-state functional connectivity and network features to classify MDD vs. healthy subjects. For each subject, mean time-series are extracted from 85 brain regions and they are decomposed to 4-frequency bands. Mean time-series for each of the frequency bands are utilized to compute the Pearson correlation and network characteristics. Features are selected separately from correlation and network characteristics using Minimum Redundancy Maximum Relevance (mRMR) to create the final classifier. The proposed scheme achieves 79% accuracy (65 out of 82 subjects classified correctly) with 86% sensitivity (42 out of 49 MDD subjects identified correctly) and 70% specificity (23 out of 33 controls identified correctly) using leave-one-out classification with in-fold feature selection. Pearson correlation had the highest discrimination in band 0.015-0.03 Hz and network based features had the highest discrimination in band 0.03-0.06 Hz for distinguishing MDD vs. healthy subjects.

Diffusion tensor imaging (DTI) has recently been added to the large scale of studies for Alzheimer's Disease (AD) to investigate the White Matter (WM) defects that are not detectable using structural MRI. In this paper, we extracted Speeded Up Robust Features (SURF) and Scale Invariant Feature Transform (SIFT) features, based on the visual diffusion patterns of Fractional Anisotropy (FA), and Mean Diffusivity (MD) maps, to build bag-of-words AD-signature for the hippocampal area. The experiments were accomplished with a subset of participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset formed of AD patients (n = 35), Early Mild Cognitive Impairment (EMCI) (n=6), Late Mild Cognitive Impairment (LMCI) (n=24) and cognitively healthy elderly Normal Controls (NC) (n=31). The preliminary studied experiments give promising results that would consider the proposed system as an accurate and useful tool to capture the AD leanness with accuracy of 87% and 89% for FA and MD maps respectively.

Proper training of convolutional neural networks (CNNs) requires annotated training datasets oflarge size, which are not currently available in CT colonography (CTC). In this paper, we propose a well-designed framework to address the challenging problem of data shortage in the training of 3D CNN for the detection of polyp candidates, which is the first and crucial part of the computer-aided diagnosis (CAD) of CTC. Our scheme relies on the following two aspects to reduce overfitting: 1) mass data augmentation, and 2) a flat 3D residual fully convolutional network (FCN). In the first aspect, we utilize extensive rotation, translation, and scaling with continuous value to provide numerous data samples. In the second aspect, we adapt the well-known V-Net to a flat residual FCN to resolve the problem of detection other than segmentation. Our proposed framework does not rely on accurate colon segmentation nor any electrical cleansing of tagged fluid, and experimental results show that it can still achieve high sensitivity with much fewer false positives. Code has been made available at: http://github.com/chenyzstju/ctc_screening_cnn.

In hospitals, physicians diagnose brain-related disorders such as epilepsy by analyzing electroencephalograms (EEG). However, manual analysis of EEG data requires highly trained clinicians or neurophysiologists and is a procedure that is known to have relatively low inter-rater agreement (IRA). Moreover, the volume of the data and rate at which new data is acquired makes interpretation a time-consuming, resource hungry, and expensive process. In contrast, automated analysis offers the potential to improve the quality of patient care by shortening the time to diagnosis, reducing manual error, and automatically detecting debilitating events. In this paper, we focus on one of the early decisions made in this process which is identifying whether an EEG session is normal or abnormal. Unlike previous approaches, we do not extract hand-engineered features but employ deep neural networks that automatically learn meaningful representations. We undertake a holistic study by exploring various pre-processing techniques and machine learning algorithms for addressing this problem and compare their performance. We have used the recently released "TUH Abnormal EEG Corpus" dataset for evaluating the performance of these algorithms. We show that modern deep gated recurrent neural networks achieve 3.47% better performance than previously reported results.

Accurate optic disc (OD) segmentation is a fundamental step in computer-aided ocular disease diagnosis. In this paper, we propose a new pipeline to segment OD from retinal fundus images based on deep object detection networks. The fundus image segmentation problem is redefined as a relatively more straightforward object detection task. This then allows us to determine the OD boundary simply by transforming the predicted bounding box into a vertical and non-rotated ellipse. Using Faster R-CNN as the object detector, our method achieves state-of-the-art OD segmentation results on ORIGA dataset, outperforming existing methods in this field.

Consolidation of the lung is a common pathology which is sometimes life threatening. One of the primary causes is the infection of lung tissue (pneumonia). Vocal resonance and vocal fremitus are part of the routine clinical examination of the respiratory system by physicians, which although time consuming, helps in the diagnosis of consolidation of the lung. In this paper, we suggest a possible automatic lung consolidation detection system that can be used by health workers with basic training. Analysis is performed on the vocal resonance sound signals recorded using an electronic stethoscope from the chest walls of normal subjects and patients. We show that for the detection system, use of signal loudness, which is generally considered by physicians, would be infeasible and propose that signal power spectral density computed using our system be considered. Certain frequency regions in the power spectral density (periodogram) are proved to be significant indicators of lung consolidation using t-tests. These findings are then applied to design the detection system using Gaussian naive Bayes classifiers.

In this paper, we propose a new secure content based image retrieval (SCBIR) system adapted to the cloud framework. This solution allows a physician to retrieve images of similar content within an outsourced and encrypted image database, without decrypting them. Contrarily to actual CBIR approaches in the encrypted domain, the originality of the proposed scheme stands on the fact that the features extracted from the encrypted images are themselves encrypted. This is achieved by means of homomorphic encryption and two non-colluding servers, we however both consider as honest but curious. In that way an end to end secure CBIR process is ensured. Experimental results carried out on a diabetic retinopathy database encrypted with the Paillier cryptosystem indicate that our SCBIR achieves retrieval performance as good as if images were processed in their non-encrypted form.

Automatic and reliable diagnosis of skin cancer, as a smartphone application, is of great interest. Among different types of skin cancers, melanoma is the most dangerous one which causes most deaths. Meanwhile, melanoma is curable if it were diagnosed in its early stages. In this paper we propose an efficient system for prescreening of pigmented skin lesions for malignancy using general-purpose digital cameras. These images can be captured by a smartphone or a digital camera. This could be beneficial in different applications, such as computer aided diagnosis and telemedicine applications. It could assist dermatologists, or smartphone users, evaluate risk of suspicious moles. The proposed method enhances borders and extracts a broad set of dermatologically important features. These discriminative features allow classification of lesions into two groups of melanoma and benign. This method is computationally appropriate as a smartphone application. Experimental results show that our proposed method is superior in diagnosis accuracy compared to state-of-the-art methods.

Depression is an increasingly prevalent mood disorder. This is the reason why the field of computer-based depression assessment has been gaining the attention of the research community during the past couple of years. The present work proposes two algorithms for depression detection, one Frame-based and the second Video-based, both employing Curvelet transform and Local Binary Patterns. The main advantage of these methods is that they have significantly lower computational requirements, as the extracted features are of very low dimensionality. This is achieved by modifying the previously proposed algorithm which considers Three-Orthogonal-Planes, to only Pairwise-Orthogonal-Planes. Performance of the algorithms was tested on the benchmark dataset provided by the Audio/Visual Emotion Challenge 2014, with the person-specific system achieving 97.6% classification accuracy, and the person-independed one yielding promising preliminary results of 74.5% accuracy. The paper concludes with open issues, proposed solutions, and future plans.

Melanoma, most threatening type of skin cancer, is on the rise. In this paper an implementation of a deep-learning system on a computer server, equipped with graphic processing unit (GPU), is proposed for detection of melanoma lesions. Clinical (non-dermoscopic) images are used in the proposed system, which could assist a dermatologist in early diagnosis of this type of skin cancer. In the proposed system, input clinical images, which could contain illumination and noise effects, are preprocessed in order to reduce such artifacts. Afterward, the enhanced images are fed to a pre-trained convolutional neural network (CNN) which is a member of deep learning models. The CNN classifier, which is trained by large number of training samples, distinguishes between melanoma and benign cases. Experimental results show that the proposed method is superior in terms of diagnostic accuracy in comparison with the state-of-the-art methods.

Oral cancer is one of the most prevalent form of cancer and its severity is aggrandized specially among the socio-economically backward population in developing countries. A major fraction of patient population is unable to avail diagnosis for oral cancer due to scarcity of state-of-the-art infrastructure and experienced oral and maxillofacial pathologist. Contemporary gold standard of oral cancer confirmation relies on biopsy report. But biopsy is invasive and thus patients are usually reluctant to undergo this test. Moreover, biopsy yields considerable false negatives if investigated tissue is not collected precisely from the carcinogenic location. Till date, there is dearth of computer aided pre-screening tool for detection of oral cancer. The paper presents Digital Infrared Thermal Imaging as a viable modality for early screening of oral cancer. This is the pioneering attempt to discriminate normal subjects from patients by leveraging discriminating texture features on oral thermograms. Statistically significant texture features were selected from a) both halves of frontal face and b) right and left profile faces. Due to disparity of distribution of facial temperature between normal subjects and patients, the corresponding texture features form discriminative class specific local clusters. Such local conglomeration was exploited using k-means and fuzzy k-means clustering. We adopt the concept of cluster prototype classifier which assigns label to each cluster according to majority class labels within that cluster. Highest classification accuracy of 86.12% is attained on fusion of features from left and right half of frontal face of precancerous subject followed by fuzzy k-means guided cluster prototype classification. The proposed work outperforms our previously developed pre-screening framework by upto 6.5%. Such promising results boosts the viability of our approach.

Computer Aided-Diagnosis (CAD) systems have been proposed to help dermatologists diagnose melanomas. However, these systems fail to provide a medical explanation for the diagnosis. This makes the dermatologists unsure about their use, since they are not easy to understand. In this paper we propose a CAD system that extracts a clinically inspired color description of the lesion and then, uses this information to discriminate melanomas from benign lesions. The proposed system is also capable of showing the extracted color features, making the system and its decisions more comprehensible for practitioners. The development of this system is hampered by the lack of a database of detailed annotate dermoscopy images. Nonetheless, we are able to tackle this issue using an image annotation framework based on the Correspondence-LDA algorithm. This method is applied with success to the identification of relevant colors in dermoscopy images, obtaining an average Precision of 84.9% and a Recall of 85.5%. The proposed color representation is then used to classify skin lesions, resulting in a Sensitivity of 78.9% and Specificity of 76.7%, these values are promising and comparable with the state-of-the art.

This paper presents a conformable wireless patch and its mobile application for physical activity, spO2 and pCO2 recording associated to digital biomarkers that aim at providing the clinicians with a reliable computer-aided diagnosis tool for rapid and continuous monitoring of sleep respiratory disorders. Each part of the system is described and results are presented and discussed. The reflectance sp02 sensor has been tested in vivo on several body sites and several subjects then compared to a reference device. The electrochemical tcpO2 sensor has been validated in vitro. Based on these physiological parameters, the proposed algorithms to automatically identifying sleep respiratory events are compared to a reference index.

Breast cancer is one of the leading cancers worldwide. Precision medicine is a new trend that systematically examines molecular and functional genomic information within each patient's cancer to identify the patterns that may affect treatment decisions and potential outcomes. As a part of precision medicine, computer-aided diagnosis enables joint analysis of functional genomic information and image from pathological images. In this paper we propose an integrated framework for breast cancer staging using image-omics and functional genomic information. The entire biomedical imaging informatics framework consists of image-omics extraction, feature combination, and classification. First, a robust automatic nuclei detection and segmentation is presented to identify tumor regions, delineate nuclei boundaries and calculate a set of image-based morphological features; next, the low dimensional image-omics is obtained through principal component analysis and is concatenated with the functional genomic features identified by a linear model. A support vector machine for differentiating stage I breast cancer from other stages are learned. We experimentally demonstrate that compared with a single type of representation (image-omics), the combination of image-omics and functional genomic feature can improve the classification accuracy by 3%.

Malignant melanoma, which is the most dangerous type of skin cancer, is commonly diagnosed in all people, regardless of age, gender, or race. In the last several years an increasing melanoma incidence and mortality rate has been observed worldwide. In this research we present a new approach to the detection and classification of border irregularity, one of the major parameter in a widely used diagnostic algorithm ABCD rule of dermoscopy. Accurate assessment of irregular borders is clinically important due to a significantly different occurrence in benign and malignant skin lesions. In this paper we describe a complex algorithm containing following steps: image enhancement, lesion segmentation, border irregularity detection as well as classification. The algorithm has been tested on 300 dermoscopic images and achieved a detection of 79% and classification accuracy of 90%. Compared to state-of-the-art, we obtain improved classification accuracy.

One of the most important signs of systemic disease that presents on the retina is vascular abnormalities such as in hypertensive retinopathy. Manual analysis of fundus images by human readers is qualitative and lacks in accuracy, consistency and repeatability. Present semi-automatic methods for vascular evaluation are reported to increase accuracy and reduce reader variability, but require extensive reader interaction; thus limiting the software-aided efficiency. Automation thus holds a twofold promise. First, decrease variability while increasing accuracy, and second, increasing the efficiency. In this paper we propose fully automated software as a second reader system for comprehensive assessment of retinal vasculature; which aids the readers in the quantitative characterization of vessel abnormalities in fundus images. This system provides the reader with objective measures of vascular morphology such as tortuosity, branching angles, as well as highlights of areas with abnormalities such as artery-venous nicking, copper and silver wiring, and retinal emboli; in order for the reader to make a final screening decision. To test the efficacy of our system, we evaluated the change in performance of a newly certified retinal reader when grading a set of 40 color fundus images with and without the assistance of the software. The results demonstrated an improvement in reader's performance with the software assistance, in terms of accuracy of detection of vessel abnormalities, determination of retinopathy, and reading time. This system enables the reader in making computer-assisted vasculature assessment with high accuracy and consistency, at a reduced reading time.

Computer-aided detection systems can help radiologists to detect pulmonary nodules at an early stage. In this paper, a novel Computer-Aided Diagnosis system (CAD) is proposed for the classification of pulmonary nodules as malignant and benign. The proposed CAD system using ensemble learning classifiers, provides an important support to radiologists at the diagnosis process of the disease, achieves high classification performance. The proposed approach with bagging classifier results in 94.7 %, 90.0 % and 77.8 % classification sensitivities for benign, malignant and undetermined classes (89.5 % accuracy), respectively.

This article reports a robust technique for extracting time plane features of Electrocardiogram (ECG) from digital images of ECG paper strips. We concluded this article reporting performance evaluation of the system developed for machine aided cardiac disease detection. Mostly paper based ECG recordings are used in developing countries and digital photographs of different leads could easily be taken and sent with a mediocre cellular phone set. Apart from extracting the features, the proposed system detects cardiac axis deviation and diagnose if Left or Right Bundle Branch Blockage (LBBB or RBBB) is present while fed with the digital photographs of different leads of ECG strips. Preprocessing of the low-resolution images involves background grid line noise removal, adaptive image binarization by Sauvola's method and Bresenham's line joining algorithm to link the ECG signature, if broken. Pattern extraction mainly delineate the time plane features like P wave, QRS complex and T wave using water reservoir based pattern recognition techniques and Discrete Wavelet Transform (DWT). Cardiac axis deviation detection is done by checking the overall voltage levels of QRS complexes of lead I, II and III. Having the knowledge of cardiac axis completes the requirements to comment on the cardiac blockage like Left or Right Bundle Branch Blockage (LBBB or RBBB). Thus, the proposed algorithm is primarily developed for machine aided diagnosis of LBBB or RBBB from the digital photographs of ECG paper strips.

The next generation of medical technology applications for hand-held portable platforms will provide a core change in performance and sophistication, transforming the way health care professionals interact with patients. This advance is particularly apparent in the delivery of cognitive patient assessments, where smartphones and tablet computers are being used to assess complex neurological conditions to provide objective, accurate and reproducible test results. This paper reports on two such applications (apps) that have been developed to assist healthcare professionals with the detection and diagnosis of dementia and delirium.

In this paper, we introduce a Concept Graph Engine (CG-Engine) that generates patient specific personalized disease ranking based on the laboratory test data. CG-Engine uses the Unified Medical Language System database as medical knowledge base. The CG-Engine consists of two concepts namely, a concept graph and its attributes. The concept graph is a two level tree that starts at a laboratory test root node and ends at a disease node. The attributes of concept graph are: Relation types, Semantic types, Number of Sources and Symmetric Information between nodes. These attributes are used to compute the weight between laboratory tests and diseases. The personalized disease ranking is created by aggregating the weights of all the paths connecting between a particular disease and contributing abnormal laboratory tests. The clinical application of CG-Engine improves physician's throughput as it provides the snapshot view of abnormal laboratory tests as well as a personalized disease ranking.

Capsule endoscopy is a patient-friendly endoscopy broadly utilized in gastrointestinal examination. However, the efficacy of diagnosis is restricted by the large quantity of images. This paper presents a modified anomaly detection method, by which both known and unknown anomalies in capsule endoscopy images of small intestine are expected to be detected. To achieve this goal, this paper introduces feature extraction using a non-linear color conversion and Higher-order Local Auto Correlation (HLAC) Features, and makes use of image partition and subspace method for anomaly detection. Experiments are implemented among several major anomalies with combinations of proposed techniques. As the result, the proposed method achieved 91.7% and 100% detection accuracy for swelling and bleeding respectively, so that the effectiveness of proposed method is demonstrated.

This paper describes a computer-aided diagnosis (CAD) method to classify diffuse lung diseases (DLD) patterns on HRCT images. Due to the high variety and complexity of DLD patterns, the performance of conventional methods on recognizing DLD patterns featured by geometrical information is limited. In this paper, we introduced a sparse representation based method to classify normal tissues and five types of DLD patterns including consolidation, ground-glass opacity, honeycombing, emphysema and nodular. Both CT values and eigenvalues of Hessian matrices were adopted to calculate local features. The 2360 VOIs from 117 subjects were separated into two independent set. One set was used to optimize parameters, and the other set was adopted to evaluation. The proposed technique has a overall accuracy of 95.4%. Experimental results show that our method would be useful to classify DLD patterns on HRCT images.

This paper investigates the combination of multiresolution methods for feature extraction for lung cancer. The focus is on the impact of combining wavelet and curvelet on the accuracy of the disease diagnosis. The paper investigates feature extraction with two different levels of wavelet, two different wavelet functions and the combination of wavelet and curvelet to obtain a high classification rate. The findings suggest the potential of combining different multiresolution methods in achieving high accuracy rates.

Thermal infrared imaging has been shown to be useful for diagnosing breast cancer, since it is able to detect small tumors and hence can lead to earlier diagnosis. In this paper, we present a computer-aided diagnosis approach for analysing breast thermograms. We extract image features that describe bilateral differences of the breast regions in the thermogram, and then feed these features to an ensemble classifier. For the classification, we present an extension to the Under-Sampling Balanced Ensemble (USBE) algorithm. USBE addresses the problem of imbalanced class distribution that is common in medical decision making by training different classifiers on different subspaces, where each subspace is created so as to resemble a balanced classification problem. To combine the individual classifiers, we use a neural fuser based on discriminants and apply a classifier selection procedure based on a pairwise double-fault diversity measure to discard irrelevant and similar classifiers. We demonstrate that our approach works well, and that it statistically outperforms various other ensemble approaches including the original USBE algorithm.

The increasing incidence of melanoma has recently promoted the development of computer-aided diagnosis systems for the classification of dermoscopic images. Unfortunately, the performance of such systems cannot be compared since they are evaluated in different sets of images by their authors and there are no public databases available to perform a fair evaluation of multiple systems. In this paper, a dermoscopic image database, called PH², is presented. The PH² database includes the manual segmentation, the clinical diagnosis, and the identification of several dermoscopic structures, performed by expert dermatologists, in a set of 200 dermoscopic images. The PH² database will be made freely available for research and benchmarking purposes.

This paper examines the feasibility of the trend covariability between QT and RR Intervals (QTIs and RRIs) be a novel mean of the sudden cardiac death (SCD) risk stratification. Twenty four hour beat to beat QTIs and RRIs are measured from Holter ECG recordings of 25 normal control subjects (SCD-C), 14 low SCD risk patients (SCD-L) with high blood pressure or light cardiac arrhythmia and 11 SCD high risk patients (SCD-H) with heart attack history. The Kalman filtering technique has been applied to decompose 24 hour short term mean QTIs and RRIs sequences into trend components and additive random variations. The correlation coefficients (TC-QT/RR) and cross entropies (TE-QT/RR) between the QT and RR trend signals are estimated. Cross entropy TE-QT/RR achieved the best stratification of subject groups. TE-QT/RR distribution for SCD-C, -L -H subject groups were 1.697 ± 0.058, 1.160 ± 0.099, 0.920 ± 0.067. The differences in entropy values are statistically significant for all classes pairs (SCD-H and -C (p<0.00001); -L and -C (p<0.001); -H and -L (p<0.05) The result indicates that the TE-QT/RR could be a novel index for the SCD risk stratification.

Diabetes is the seventh leading cause of death in the United States. In 2010, about 1.9 million new cases of diabetes were diagnosed in people aged 20 years or older. Remote health monitoring systems can help diabetics and their healthcare professionals monitor health-related measurements by providing real-time feedback. However, data-driven methods to dynamically prioritize and generate tasks are not well investigated in the remote health monitoring. This paper presents a task optimization technique used in WANDA (Weight and Activity with Blood Pressure and Other Vital Signs); a wireless health project that leverages sensor technology and wireless communication to monitor the health status of patients with diabetes. WANDA applies data analytics in real-time to improving the quality of care. The developed algorithm minimizes the number of daily tasks required by diabetic patients using association rules that satisfies a minimum support threshold. Each of these tasks maximizes information gain, thereby improving the overall level of care. Experimental results show that the developed algorithm can reduce the number of tasks up to 28.6% with minimum support 0.95, minimum confidence 0.97 and high efficiency.

Intermittent Claudication is a walking symptom. After a short time walking, patients suffer from pains at lower limbs. But if taking a rest, the pains can be relieved and they can walk again. Unfortunately, it arises from not one but mainly two kinds of diseases: LSS (lumber spinal canal stenosis) and PAD (peripheral arterial disease). Additionally, it is reported that symptom is similar and LSS groups is furthermore divided into two main groups: L4 and L5 groups. Therefore, it is clinically very important to differentiate which diseases the patients suffer from, PAD, L4 or L5. We aims at developing the system to differentiate them from short walking motion data. In our previous paper, we derived differentiation factors, but did not consider the difference between L4 and L5 and the results are limited. This paper focuses on biarticular muscles associated with the diseases, and derive new and effective differentiation factors. The results supports their effectiveness and validity.

One of the obstacles in the development of rehabilitation robotics has been inadequacy in the measurement of treatment effects due to interventions. A measurement tool that will efficiently produce a large reliable sample of measurements collected during a single session that can also produce a rich set of data which reflects a subject's ability to perform meaningful functional activities has not been developed. This paper presents three linear regression models generated from seven kinematic measures collected during the performance of virtually simulated rehabilitation activities that were integrated with haptic robots by 19 persons with upper extremity hemiparesis due to chronic stroke. One of these models demonstrated a statistically significant correlation with the subjects' scores on the Jebsen Test of Hand Function (JTHF), a battery of six standardized upper extremity functional activities. The second and third models demonstrated a statistically significant correlation with the subjects' change scores on the JTHF.

Liver steatosis is a common disease usually associated with social and genetic factors. Early detection and quantification is important since it can evolve to cirrhosis. Steatosis is usually a diffuse liver disease, since it is globally affected. However, steatosis can also be focal affecting only some foci difficult to discriminate. In both cases, steatosis is detected by laboratorial analysis and visual inspection of ultrasound images of the hepatic parenchyma. Liver biopsy is the most accurate diagnostic method but its invasive nature suggest the use of other non-invasive methods, while visual inspection of the ultrasound images is subjective and prone to error. In this paper a new Computer Aided Diagnosis (CAD) system for steatosis classification and analysis is presented, where the Bayes Factor, obatined from objective intensity and textural features extracted from US images of the liver, is computed in a local or global basis. The main goal is to provide the physician with an application to make it faster and accurate the diagnosis and quantification of steatosis, namely in a screening approach. The results showed an overall accuracy of 93.54% with a sensibility of 95.83% and 85.71% for normal and steatosis class, respectively. The proposed CAD system seemed suitable as a graphical display for steatosis classification and comparison with some of the most recent works in the literature is also presented.

This paper proposes a real time digitally assisted analog motion artifact reduction ASIC with ECG measurement simultaneously. It features one ECG monitoring and in- and quad-phase electrode-skin impedance measurement, which are used to estimate motion artifacts. The implemented ASIC is capable of actual motion artifact reduction in the analog domain before final amplification.

The relevance of psychophysiological measurements for affective computing and emotion analysis applications has been widely recognized. However, and although several authors have studied the informative content of parameters derived from cardiovascular and other modalities, feature extraction remains an open topic in the field. This is particularly relevant in scenarios where the autonomic nervous system triggering stimuli are unknown. In this paper, we analyze a set of features extracted from multimodal biosignal data, applicable to the assessment of psychophysiological load in unconstrained settings. Experimental evaluation is performed on real world data, collected both from control subjects and subjects with a strong clinical background, in a context of questionnaire-based clinical history reporting. The devised feature set has shown promising properties, making it prone to complement the more traditional measurements.

Essential tremor is a disorder that causes involuntary oscillations in patients while engaging in actions or while maintaining a posture. ET patients have serious difficulties in performing activities of daily living such as eating food, drinking water, and writing. We have thus been developing an EMG-controlled exoskeletal robot to suppress tremors. The EMG signal of ET patients involves a mix of voluntary movement and tremor signals. To control the exoskeletal robot accurately, tremor signals must be removed from the patient's EMG signal. To date, we have been developing a filter to remove tremor signals from the patient's EMG. The design of this filter was based on the hypothesis that the rectified tremor signals are able to be approximated by a powered sine wave. This filter was found to have a large effect on removing tremor signals. However, tremor signals are generated both while performing voluntary movement and while maintaining a posture, and the filter was attenuating both signals. To control this robot accurately, the signal generated while performing voluntary movement is expected not to be attenuated. To accomplish this, we try to use a parameter that reflects a state of the patient's movement, performing a voluntary movement or maintaining a posture, as a switch to activate the powered sine filter. This paper provides an analysis of the favorable parameters. We focus on two parameters: the peak-to-peak interval of the rectified EMG signal, and the interval of the flat and low amplitude area of the rectified EMG signal. Through evaluation, it is affirmed that both parameters change with the state of the patient's movement. However, the latter parameter is superior to the former in terms of variability, which indicates that the interval of the flat and low amplitude area of the rectified EMG signal is a more favorable parameter to promote control of the exoskeletal robot. As a future work, we will mount the parameter to the algorithm and evaluate the robotic system.

Constructing statistical models of electrocardiogram (ECG) signals, whose parameters can be used for automated disease classification, is of great importance in precluding manual annotation and providing prompt diagnosis of cardiac diseases. ECG signals consist of several segments with different morphologies (namely the P wave, QRS complex and the T wave) in a single heart beat, which can vary across individuals and diseases. Also, existing statistical ECG models exhibit a reliance upon obtaining a priori information from the ECG data by using preprocessing algorithms to initialize the filter parameters, or to define the user-specified model parameters. In this paper, we propose an ECG modeling technique using the sequential Markov chain Monte Carlo (SMCMC) filter that can perform simultaneous model selection, by adaptively choosing from different representations depending upon the nature of the data. Our results demonstrate the ability of the algorithm to track various types of ECG morphologies, including intermittently occurring ECG beats. In addition, we use the estimated model parameters as the feature set to classify between ECG signals with normal sinus rhythm and four different types of arrhythmia.

Knowing how well an activity is performed is important for home rehabilitation. We would like to not only know if a motion is being performed correctly, but also in what way the motion is incorrect so that we may provide feedback to the user. This paper describes methods for assessing human motion quality using body-worn tri-axial accelerometers and gyroscopes. We use multi-label classifiers to detect subtle errors in exercise performances of eight individuals with knee osteoarthritis, a degenerative disease of the cartilage. We present results obtained using various machine learning methods with decision tree base classifiers. The classifier can detect classes in multi-label data with 75% sensitivity, 90% specificity and 80% accuracy. The methods presented here form the basis for an at-home rehabilitation device that will recognize errors in patient exercise performance, provide appropriate feedback on the performance, and motivate the patient to continue the prescribed regimen.

Life signals from human body, e.g. heartbeat or electrocardiography (ECG), are usually weak and susceptible to external noise and interference. Adaptive filter is a good tool to reduce the influence of ambient noise/interference on the life signals. Least mean squares (LMS) algorithm, as one of most popular adaptive algorithms for active noise cancellation (ANC) by adaptive filtering, has the advantage of easy implementation. In order to further decrease the complexity of LMS algorithm based adaptive filter, a Log-LMS algorithm was proposed, which quantized signals by the function of log2. The algorithm can replace multipliers by simple shifting. However, both LMS algorithm and Log-LMS algorithm have the disadvantage of serious signal distortion in biomedical applications. In this paper, a modified Log-LMS algorithm is presented, which divides the convergence process into two different stages, and utilizes different quantization method in each stage. Two scenarios of biomedical applications are used for analysis, 1) using stethoscope in emergence medical helicopter and 2) measuring ECG under power line interference. The simulated results show that the modified algorithm can achieve fast convergence and low signal distortion in processing periodic life signals.

Several clinical studies have reported that EEG synchrony is affected by Alzheimer's disease (AD). In this paper a frequency band analysis of AD EEG signals is presented, with the aim of improving the diagnosis of AD using EEG signals. In this paper, multiple synchrony measures are assessed through statistical tests (Mann-Whitney U test), including correlation, phase synchrony and Granger causality measures. Moreover, linear discriminant analysis (LDA) is conducted with those synchrony measures as features. For the data set at hand, the frequency range (5-6 Hz) yields the best accuracy for diagnosing AD, which lies within the classical theta band (4-8 Hz). The corresponding classification error is 4.88% for directed transfer function (DTF) Granger causality measure. Interestingly, results show that EEG of AD patients is more synchronous than in healthy subjects within the optimized range 5-6 Hz, which is in sharp contrast with the loss of synchrony in AD EEG reported in many earlier studies. This new finding may provide new insights about the neurophysiology of AD. Additional testing on larger AD datasets is required to verify the effectiveness of the proposed approach.

Knee joint vibration signals are very useful for computer-aided analysis of the pathological conditions in the knee. In a vibration arthrometry test, the legs of patients with knee joint disorders may tremble due to the reaction of pain, which causes the baseline wander that may affect the diagnostic decision making in medical study. This paper presents a new type of cascade moving average filter with hierarchical layers to remove the baseline wander in the raw knee joint vibration signals. The first layer of the cascade filter contains two moving averaging operators with the same order. The five tail inputs of the first moving averaging operator are overlapping with the beginning inputs of the successive operator. The piecewise linear trends estimated by the moving average operators in the first layer were smoothed in the final cascade filter output. The simulation results showed that the cascade filter can effectively remove the baseline wander in the raw knee joint vibration signals.

This paper presents a recurrence network approach to quantify dynamic complexity of skin blood flow oscillations (BFO) in response to loading pressure. This approach consists of three processes, including 1) phase space reconstruction by means of time delay embedding, 2) construction of a recurrence matrix that represents neighboring states in phase space, and 3) consideration of the recurrence matrix as an adjacency matrix representing links in a network and the use of clustering coefficients to characterize phase space properties. By using the Lorenz system and real data, we demonstrate that the global clustering coefficient is robust to the embedding parameters. We applied this approach to study skin BFO at baseline and during loading pressure, a causative factor of skin breakdown. The results showed that global clustering coefficients of BFO significantly decreased in response to loading ( &#60;0.05). Moreover, surrogate tests indicated that such a decrease was associated with a loss of nonlinearity of BFO. Our results suggest that the recurrence network approach can practically quantify the nonlinear dynamics of BFO.

Heart sound localization in chest sound is an essential part for many heart sound cancellation algorithms. The main difficulty for heart sound localization methods is the precise determination of the onset and offset boundaries of the heart sound segment. This paper presents a novel method to estimate lower and upper bounds for the onset and offset of the heart sound segment, which can be used as anchor points for more precise estimation. For this purpose, first chest sound is divided into frames and then entropy and smoothed entropy features of these frames are extracted, and used in the Convex-hull algorithm to estimate the upper and lower bounds for heart sound boundaries. The Convex-hull algorithm constructs a special type of envelope function for entropy features and if the maximal difference between the envelope function and the entropy is larger than a certain threshold, this point is considered as a heart sound bound. The results of the proposed method are compared with a baseline method which is a modified version of a well-known heart sound localization method. The results show that the proposed method outperforms the baseline method in terms of accuracy and detection error rate. Also, the experimental results show that smoothing entropy features significantly improves the performance of both baseline and proposed methods.

This paper describes the development, deployment and trial results from 9 volunteers using the eCAALYX system. The eCAALYX system is an ambient assisted living telemonitoring system aimed at older adults suffering with co-morbidity. Described is a raw account of the challenges that exist and results in bringing a Telemedicine system from laboratory to real-world implementation and results for usability, functionality and reliability.

In this paper, we report an experimental comparison of dynamic support vector machines (SVMs) to dynamic neural networks (DNNs) in the context of a system for detecting dyskinesia and tremor in Parkinson's disease (PD) patients wearing accelerometer (ACC) and surface electromyographic (sEMG) sensors while performing unscripted and unconstrained activities of daily living. These results indicate that SVMs and DNNs of comparable computational complexities yield approximately identical performance levels when using an identical set of input features.

In this paper a novel Lung Sound Automatic Verification (LSAV) system and front-end Quantile based acoustic models to classify Lung Sounds (LS) are proposed. The utilization of Quantiles allowed an easier and objective assessment with smaller computational demand. Moreover, less-complex Gaussian Mixture Models (GMM) were computed than those previously reported. The LSAV system allowed us to reach practically negligible error in healthy (normal) LS verification. LASV system efficiency and the optimal GMM's were evaluated by using Equal Error Rate (EER) and Bayesian Information Criterion (BIC) techniques respectively. These approaches could provide a tool for broader medical evaluation which does not rely, as it is often the case, on a qualitative and subjective description of LS.

Chronic Obstructive Pulmonary Disease (COPD) is a major cause of morbidity and mortality in Australia and globally, and leads to a substantial burden on healthcare services. Effective and timely management of patients with COPD has been essential to alleviate COPD exacerbation, improve the quality of life, and consequently reduce the economic burden. To achieve this, a mobile and internet technologies assisted home care model (M-COPD) was developed to assist clinicians to remotely monitor and manage COPD conditions and events. This paper will focus on the technical aspect of M-COPD system by describing its setup and discussing how the M-COPD could address the clinical needs in monitoring and managing COPD conditions of patients at home.

Wrist actigraphy is a well established procedure to monitor human activity. Among other areas, it has a special relevance in sleep studies where its lightweight and non-intrusive nature make it a valuable tool to access the circadian cycle. While there are several methods to extract information from the data, the differentiation between sleep and wakefulness states is still an open discussion. In this paper, the characteristics of the movements in the different states are assumed to be intrinsically different. These differences are not simply related with magnitude and movement counting, but due to real differences on the statistical distributions describing the actigraphy data. Thus it is possible to refine the discrimination level when detecting these states. The proposed methodology to characterize the actigraphy data is based on a mixture of three canonical distributions; i)Exponential, ii)Rayleigh and iii)Gaussian. It is shown that the weights and parameters estimated in each state are organized into almost separable clusters on the feature space. This suggests the ability of the method to discriminate these states based only on the movements recorded on actigraphy data.

Intracranial electroencephalograms (EEG) provide a direct observation of neural activity by placing an electrode array on the cortical surface near the suspected epileptic foci. The neural spikes appeared during inter-ictal stages are mainly produced by abnormal neural discharges from epileptic foci. The topological mapping of spikes' potentials is commonly used to identify the epileptogenic zone. However, the propagations among multi-channel spikes are also important to identify the epileptic source activity. In addition, the changes of source activities in a series of consecutive spikes reveal the time-varying neural activations during discharge process, which provide alternative information for interpreting epileptic source activity. This paper proposes a spike classification based on the similarity of phase-space features to select candidate spikes from the intracranial EEGs recorded from an 8×8 electrocorticogram grid. Then, the partial directed coherence (PDC), which can provide the flow of source activity, at each spiking time point is computed. The outflow PDCs of all electrodes are therefore displayed on the grid. Our result showed that the derived source activities in the preceding spikes had high concentrated distributions but decreased in latter spikes. This implied the epileptic discharges were initially induced by a small-area cortex neurons and then spread out.

Automatic sleep staging from convenient and unobtrusive sensors has received considerable attention lately because this can enable a large range of potential applications in the clinical and consumer fields. In this paper the focus is on achieving non-REM (NREM) sleep staging from ocular electrodes. From these signals, specific patterns related to sleep such as slow eye movements, K-complexes, eye blinks, and spectral features are estimated. Although such patterns are characteristic of the Electroencephalogram, they can also be visible to a lesser extent on signals from ocular electrodes. Automatic sleep staging was implemented using two approaches: i) based on a state-machine and ii) using a neural network. The first one relied on the recommendations of the American Academy of Sleep Medicine, and the second one used a multilayer perceptron which was trained on manually sleep-staged data. Results were obtained on the data of five volunteers who participated in a nap experiment. Manual sleep staging of this data, performed by an expert, was used as reference. Five stages were considered, namely wake with eyes open, wake with eyes closed, and sleep stages N1, N2, and N3. The results were characterized in terms of confusion matrices from which the Cohen's κ coefficients were estimated. The values of κ for both the state-machine and neural-network based automatic sleep staging approaches were 0.79 and 0.59 respectively. Thus, the state-machine based approach shows a very good agreement with manual staging of sleep-data.

With the advent of low-cost, high-fidelity, and long lasting sensors in recent years, it has become possible to acquire biomedical signals cheaply and remotely over a prolonged period of time. Oftentimes different types of sensors are deployed in the hope of capturing precursor patterns that are highly correlated to a particular clinical episode, such as seizure, congestive heart failure etc. While there have been several studies that successfully identify patterns as reliable precursors for specific medical conditions, most of them require domain-specific knowledge and expertise. The developed algorithms are also unlikely to be applicable to other medical conditions. In this paper we present a generalized algorithm that discovers potential precursor patterns without prior knowledge or domain expertise. The algorithm makes use of wavelet transform and information theory to extract generic features, and it is also classifier agnostic. Based on experiment results using three distinct datasets collected from real-world patients, our algorithm has attained performance comparable to those obtained from previous studies that rely heavily on domain-expert knowledge. Furthermore, the algorithm also discovers non-trivial knowledge in the process.