Efficient anomaly detection from medical signals and images with convolutional neural networks for Internet of medical things (IoMT) systems.

Deep learning is one of the most promising machine learning techniques that revolutionalized the artificial intelligence field. The known traditional and convolutional neural networks (CNNs) have been utilized in medical pattern recognition applications that depend on deep learning concepts. This is attributed to the importance of anomaly detection (AD) in automatic diagnosis systems. In this paper, the AD is performed on medical electroencephalography (EEG) signal spectrograms and medical corneal images for Internet of medical things (IoMT) systems. Deep learning based on the CNN models is employed for this task with training and testing phases. Each input image passes through a series of convolution layers with different kernel filters. For the classification task, pooling and fully-connected layers are utilized. Computer simulation experiments reveal the success and superiority of the proposed models for automated medical diagnosis in IoMT systems.

A Comprehensive Evaluation and Benchmarking of Convolutional Neural Networks for Melanoma Diagnosis.

Melanoma is the most invasive skin cancer with the highest risk of death. While it is a serious skin cancer, it is highly curable if detected early. Melanoma diagnosis is difficult, even for experienced dermatologists, due to the wide range of morphologies in skin lesions. Given the rapid development of deep learning algorithms for melanoma diagnosis, it is crucial to validate and benchmark these models, which is the main challenge of this work. This research presents a new benchmarking and selection approach based on the multi-criteria analysis method (MCDM), which integrates entropy and the preference ranking organization method for enrichment of evaluations (PROMETHEE) methods. The experimental study is carried out in four phases. Firstly, 19 convolution neural networks (CNNs) are trained and evaluated on a public dataset of 991 dermoscopic images. Secondly, to obtain the decision matrix, 10 criteria, including accuracy, classification error, precision, sensitivity, specificity, F1-score, false-positive rate, false-negative rate, Matthews correlation coefficient (MCC), and the number of parameters are established. Third, entropy and PROMETHEE methods are integrated to determine the weights of criteria and rank the models. Fourth, the proposed benchmarking framework is validated using the VIKOR method. The obtained results reveal that the ResNet101 model is selected as the optimal diagnosis model for melanoma in our case study data. Thus, the presented benchmarking framework is proven to be useful at exposing the optimal melanoma diagnosis model targeting to ease the selection process of the proper convolutional neural network architecture.

Utilization of image interpolation and fusion in brain tumor segmentation.

Brain tumor is a mass of anomalous cells in the brain. Medical imagining techniques have a vital role in the diagnosis of brain tumors. Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) techniques are the most popular techniques to localize the tumor area. Brain tumor segmentation is very important for the diagnosis of tumors. In this paper, we introduce a framework to perform brain tumor segmentation, and then localize the region of the tumor, accurately. The proposed framework begins with the fusion of MR and CT images by the Non-Sub-Sampled Shearlet Transform (NSST) with the aid of the Modified Central Force Optimization (MCFO) to get the optimum fusion result from the quality metrics perspective. After that, image interpolation is applied to obtain a High-Resolution (HR) image from the Low-Resolution (LR) ones. The objective of the interpolation process is to enrich the details of the fusion result prior to segmentation. Finally, the threshold and the watershed segmentation are applied sequentially to localize the tumor region, clearly. The proposed framework enhances the efficiency of segmentation to help the specialists diagnose brain tumors.

Classification of retinal images based on convolutional neural network.

Automatic detection of maculopathy disease is a very important step to achieve high-accuracy results for the early discovery of the disease to help ophthalmologists to treat patients. Manual detection of diabetic maculopathy needs much effort and time from ophthalmologists. Detection of exudates from retinal images is applied for the maculopathy disease diagnosis. The first proposed framework in this paper for retinal image classification begins with fuzzy preprocessing in order to improve the original image to enhance the contrast between the objects and the background. After that, image segmentation is performed through binarization of the image to extract both blood vessels and the optic disc and then remove them from the original image. A gradient process is performed on the retinal image after this removal process for discrimination between normal and abnormal cases. Histogram of the gradients is estimated, and consequently the cumulative histogram of gradients is obtained and compared with a threshold cumulative histogram at certain bins. To determine the threshold cumulative histogram, cumulative histograms of images with exudates and images without exudates are obtained and averaged for each type, and the threshold cumulative histogram is set as the average of both cumulative histograms. Certain histogram bins are selected and thresholded according to the estimated threshold cumulative histogram, and the results are used for retinal image classification. In the second framework in this paper, a Convolutional Neural Network (CNN) is utilized to classify normal and abnormal cases.

Automated monitoring of behaviour in zebrafish after invasive procedures.

Fish are used in a variety of experimental contexts often in high numbers. To maintain their welfare and ensure valid results during invasive procedures it is vital that we can detect subtle changes in behaviour that may allow us to intervene to provide pain-relief. Therefore, an automated method, the Fish Behaviour Index (FBI), was devised and used for testing the impact of laboratory procedures and efficacy of analgesic drugs in the model species, the zebrafish. Cameras with tracking software were used to visually track and quantify female zebrafish behaviour in real time after a number of laboratory procedures including fin clipping, PIT tagging, and nociceptor excitation via injection of acetic acid subcutaneously. The FBI was derived from activity and distance swum measured before and after these procedures compared with control and sham groups. Further, the efficacy of a range of drugs with analgesic properties to identify efficacy of these agents was explored. Lidocaine (5 mg/L), flunixin (8 mg/L) and morphine (48 mg/L) prevented the associated reduction in activity and distance swum after fin clipping. From an ethical perspective, the FBI represents a significant refinement in the use of zebrafish and could be adopted across a wide range of biological disciplines.

A Smart Context-Aware Hazard Attention System to Help People with Peripheral Vision Loss.

Peripheral vision loss results in the inability to detect objects in the peripheral visual field which affects the ability to evaluate and avoid potential hazards. A different number of assistive navigation systems have been developed to help people with vision impairments using wearable and portable devices. Most of these systems are designed to search for obstacles and provide safe navigation paths for visually impaired people without any prioritisation of the degree of danger for each hazard. This paper presents a new context-aware hybrid (indoor/outdoor) hazard classification assistive technology to help people with peripheral vision loss in their navigation using computer-enabled smart glasses equipped with a wide-angle camera. Our proposed system augments users' existing healthy vision with suitable, meaningful and smart notifications to attract the user's attention to possible obstructions or hazards in their peripheral field of view. A deep learning object detector is implemented to recognise static and moving objects in real time. After detecting the objects, a Kalman Filter multi-object tracker is used to track these objects over time to determine the motion model. For each tracked object, its motion model represents its way of moving around the user. Motion features are extracted while the object is still in the user's field of vision. These features are then used to quantify the danger using five predefined hazard classes using a neural network-based classifier. The classification performance is tested on both publicly available and private datasets and the system shows promising results with up to 90% True Positive Rate (TPR) associated with as low as 7% False Positive Rate (FPR), 13% False Negative Rate (FNR) and an average testing Mean Square Error (MSE) of 8.8%. The provided hazard type is then translated into a smart notification to increase the user's cognitive perception using the healthy vision within the visual field. A participant study was conducted with a group of patients with different visual field defects to explore their feedback about the proposed system and the notification generation stage. The real-world outdoor evaluation of human subjects is planned to be performed in our near future work.

Modular design of fatigue detection in naturalistic driving environments.

Research in driver mental fatigue is motivated by the fact that errors made by drivers often have life-threatening consequences. This paper proposes a new modular design approach for the early detection of driver fatigue system taking into account optimisation of system performance using particle swarm optimisation (PSO). The proposed system is designed and implemented using an existing dataset that was simultaneously collected from participants and vehicles in a naturalistic environment. Four types of data are considered as fatigue-related metrics including: vehicle acceleration, vehicle rotation pattern, driver's head position and driver's head rotation. The driver's blink rate data is used in this work as a proxy for ground truth for the classification algorithm. The collected data elements are initially fed to input modules represented by ternary neural network classifiers that estimates alertness. A Bayesian algorithm with PSO is then used to combine and optimise detection performance based on the number of existing input modules as well as their output states. Performance of the developed fatigue-detection system is assessed experimentally with a small data samples of driver trips. The obtained results are found in agreement with the state-of-the-art in terms of accuracy (90.4%), sensitivity (92.6%) and specificity (90.7%). These results are achieved with significant design flexibility and robustness against partial loss of input data source(s). However, due to small sample size of dataset (N = 3), a larger dataset need to be tested with the same system framework to generalise the findings of this work.

Impact of stress, fear and anxiety on the nociceptive responses of larval zebrafish.

Both adult and larval zebrafish have been demonstrated to show behavioural responses to noxious stimulation but also to potentially stress- and fear or anxiety- eliciting situations. The pain or nociceptive response can be altered and modulated by these situations in adult fish through a mechanism called stress-induced analgesia. However, this phenomenon has not been described in larval fish yet. Therefore, this study explores the behavioural changes in larval zebrafish after noxious stimulation and exposure to challenges that can trigger a stress, fear or anxiety reaction. Five-day post fertilization zebrafish were exposed to either a stressor (air emersion), a predatory fear cue (alarm substance) or an anxiogenic (caffeine) alone or prior to immersion in acetic acid 0.1%. Pre- and post-stimulation behaviour (swimming velocity and time spent active) was recorded using a novel tracking software in 25 fish at once. Results show that larvae reduced both velocity and activity after exposure to the air emersion and alarm substance challenges and that these changes were attenuated using etomidate and diazepam, respectively. Exposure to acetic acid decreased velocity and activity as well, whereas air emersion and alarm substance inhibited these responses, showing no differences between pre- and post-stimulation. Therefore, we hypothesize that an antinociceptive mechanism, activated by stress and/or fear, occur in 5dpf zebrafish, which could have prevented the larvae to display the characteristic responses to pain.

Reduction in activity by noxious chemical stimulation is ameliorated by immersion in analgesic drugs in zebrafish.

Research has recently demonstrated that larval zebrafish show similar molecular responses to nociception to those of adults. Our study explored whether unprotected larval zebrafish exhibited altered behaviour after exposure to noxious chemicals and screened a range of analgesic drugs to determine their efficacy to reduce these responses. This approach aimed to validate larval zebrafish as a reliable replacement for adults as well as providing a high-throughput means of analysing behavioural responses. Zebrafish at 5 days post-fertilization were exposed to known noxious stimuli: acetic acid (0.01%, 0.1% and 0.25%) and citric acid (0.1%, 1% and 5%). The behavioural response of each was recorded and analysed using novel tracking software that measures time spent active in 25 larvae at one time. Subsequently, the efficacy of aspirin, lidocaine, morphine and flunixin as analgesics after exposure to 0.1% acetic acid was tested. Larvae exposed to 0.1% and 0.25% acetic acid spent less time active, whereas those exposed to 0.01% acetic acid and 0.1-5% citric acid showed an increase in swimming activity. Administration of 2.5 mg l-1 aspirin, 5 mg l-1 lidocaine and 48 mg l-1 morphine prevented the behavioural changes induced by acetic acid. These results suggest that larvae respond to a noxious challenge in a similar way to adult zebrafish and other vertebrates and that the effect of nociception on activity can be ameliorated by using analgesics. Therefore, adopting larval zebrafish could represent a direct replacement of a protected adult fish with a non-protected form in pain- and nociception-related research.

Decision support system for the detection and grading of hard exudates from color fundus photographs.

Diabetic retinopathy is a major cause of blindness, and its earliest signs include damage to the blood vessels and the formation of lesions in the retina. Automated detection and grading of hard exudates from the color fundus image is a critical step in the automated screening system for diabetic retinopathy. We propose novel methods for the detection and grading of hard exudates and the main retinal structures. For exudate detection, a novel approach based on coarse-to-fine strategy and a new image-splitting method are proposed with overall sensitivity of 93.2% and positive predictive value of 83.7% at the pixel level. The average sensitivity of the blood vessel detection is 85%, and the success rate of fovea localization is 100%. For exudate grading, a polar fovea coordinate system is adopted in accordance with medical criteria. Because of its competitive performance and ability to deal efficiently with images of variable quality, the proposed technique offers promising and efficient performance as part of an automated screening system for diabetic retinopathy.

Automated localization of retinal features.

Retinal fundus images are widely used in the diagnosis and treatment of various eye diseases, such as diabetic retinopathy and glaucoma. A computer-aided retinal fundus image analysis could provide an immediate detection and characterization of retinal features prior to specialist inspection. This paper proposes an approach to automatically localize the main features in fundus images, such as blood vessels, optic disc, and fovea by exploiting the spatial and geometric relations that govern their distribution within the fundus image. The blood vessels are segmented by scale-space analysis. The average thickness of these blood vessels is then computed using the vessels centerlines and orientations from a Hessian matrix. The optic disc is localized using the circular Hough transform, the parabolic Hough transform fitting, and the localization of the fovea. The proposed method can be extended to establish a foveal coordinate system to facilitate grading lesions based on the spatial relationships between lesions and landmark features. The proposed method was evaluated on publicly available image databases, and the results have demonstrated a significant improvement over the current state-of-the-art methods.

A mechatronic valve in the management of hydrocephalus: methods and performance.

The problem of excess cerebrospinal fluid in the brain (hydrocephalus) is generally managed using a passive pressure or flow regulated mechanical shunt. Despite the success of such devices, they have been plagued with a number of problems. It is desirable to have a shunt valve that responds dynamically to the changing needs of the patient, opening and closing according to a dynamic physiological pattern, rather than simply to the hydrostatic pressure across the valve. Such a valve would by necessity be mechatronic, electronically controlled by software. In this article, different methods for controlling such a mechatronic valve are explored, and the effect of current hydrocephalus management techniques on the intracranial hydrodynamics of acute hydrocephalus patient compared with those based on a mechatronic valve was investigated using numerical simulation. Furthermore, the performance of these techniques was evaluated based on a proposed multi-dimensional figure of merit. In addition, an empirical valve schedule was proposed based on different criterions. An intelligent shunting system is seen as the future in hydrocephalus management and treatment, and towards this end, suitably programmed mechatronic valves would attempt to mimic normal physiology and potentially overcome many of the problems associated with current mechanical valves.

Automated detection of red lesions from digital colour fundus photographs.

Earliest signs of diabetic retinopathy, the major cause of vision loss, are damage to the blood vessels and the formation of lesions in the retina. Early detection of diabetic retinopathy is essential for the prevention of blindness. In this paper we present a computer-aided system to automatically identify red lesions from retinal fundus photographs. After pre-processing, a morphological technique was used to segment red lesion candidates from the background and other retinal structures. Then a rule-based classifier was used to discriminate actual red lesions from artifacts. A novel method for blood vessel detection is also proposed to refine the detection of red lesions. For a standarised test set of 219 images, the proposed method can detect red lesions with a sensitivity of 89.7% and a specificity of 98.6% (at lesion level). The performance of the proposed method shows considerable promise for detection of red lesions as well as other types of lesions.

Intelligent shunt agent for gradual shunt removal.

When passive shunts, which divert cerebrospinal fluid (CSF) from the ventricles in the brain to another part of the body, were developed, apparently they change favourably the treatment of hydrocephalus, then it becomes of great importance to overcome the drawbacks of such shunts, and the gradual rising use of various shunts are accompanied by total shunt dependency with several problems and shortcomings has understandably become obvious among physicians as well as surgeons to rehabilitate and upgrade these shunts. There is a little use of carrying out arrested hydrocephalus which is subject to many aspects, ranging from problems of immediate clinical concern to the more unknowable areas of cerebrospinal fluid CSF dynamics, and it is not always as easy to define indications for arrested hydrocephalus or to evaluate the results of such treatment. However, it is important to attempt to define as precisely as possible a technique to measure the ability of arresting hydrocephalus, while current solutions estimations are based on long time procedure, evaluate parameters such as head growth, or ventricle sizes using CT or MRI scan. This paper proposes a new treatment approach and shunting system that helps improving diagnosis and treatment of Hydrocephalus patients. This approach suggests a developing and utilising an intelligent shunt agent (i-Shunt) that can learn from the patient's status and initiate a weaning program, and based on the response evaluation, the parameters of the shunt can be modified to accommodate the patient's needs. Therefore, a novel shunt could be build to satisfy the patient's need instantaneously by keeping the intracranial pressure (ICP) within normal levels, where it is actually directed toward shunt independency.

Detection of exudates in retinal images using a pure splitting technique.

Diabetic retinopathy is a major cause of blindness. Earliest signs of diabetic retinopathy are damage to blood vessels in the eye and then the formation of lesions in the retina. This paper presents an automated method for the detection of bright lesions (exudates) in retinal images. In this work, an adaptive thresholding based on a novel algorithm for pure splitting of the image is proposed. A coarse segmentation based on the calculation of a local variation for all image pixels is used to outline the boundaries of all candidates which have clear borders. A morphological operation is used to refine the adaptive thresholding results based on the coarse segmentation results. Using a clinician reference standard (ground truth), images with exudates were detected with 91.2% sensitivity, 99.3% specificity, and 99.5% accuracy. Due to its results the proposed method can achieve superior performance compared to existing techniques and is robust to image quality variability.

An expert system for hydrocephalus patient feedback.

Diagnosis of hydrocephalus symptoms and shunting system faults currently are based on clinical observation, monitoring of cranial growth, transfontanelle pressure, imaging techniques and, on occasion, studies of cerebrospinal fluid (CSF) dynamics. Up to date, the patient has to visit the hospital or meet consultant to diagnose the symptoms that occur due to rising of intracranial pressure or any shunt complications, which cause suffering for the patient and his family. This work presents the design and implementation of an expert system based on real-time patient feedback that aims to provide a suitable decision for hydrocephalus management and shunt diagnosis. Such decision would help in personalising the management as well as detecting and identifying of any shunt malfunctions without the need to contact or visit the hospital. In this paper, the development of patient feedback expert system is described. The outcome of such system would help satisfy the patient's needs regarding his/her shunt.

High resolution image acquisition from magnetic resonance and computed tomography scans using the curvelet fusion algorithm with inverse interpolation techniques.

We present a new approach, based on the curvelet transform, for the fusion of magnetic resonance and computed tomography images. The objective of this fusion process is to obtain images, with as much detail as possible, for medical diagnosis. This approach is based on the application of the additive wavelet transform on both images and the segmentation of their detail planes into small overlapping tiles. The ridgelet transform is then applied on each of these tiles, and the fusion process is performed on the ridgelet transforms of the tiles. To maximize the benefit of the fused images, inverse interpolation techniques are used to obtain high resolution images from the low resolution fused images. Three inverse interpolation techniques are presented and compared. Simulation results show the superiority of the proposed curvelet fusion approach to the traditional discrete wavelet transform fusion technique. Results also reveal that inverse interpolation techniques have succeeded in obtaining high resolution images from the fused images with better quality than that of the traditional cubic spline interpolation technique.

Instantiating a mechatronic valve schedule for a hydrocephalus shunt.

Hydrocephalus is caused by blockage or reabsorption difficulty that upsets the natural balance of production and absorption of cerebrospinal fluid in the brain, resulting in a build-up of the fluid in the ventricles of the brain. One of the recent advances in the treatment of hydrocephalus is the invention of a mechatronic valve. The desirability of such valve lies in the potential of having shunt that not only control hydrocephalus but also seeks to treat it. In contrast to current valves, such a valve is regulated based on a time based schedule not on the differential pressure across the valve. Thus the effectiveness of such valve is highly dependant on selecting an appropriate valve schedule that delivers personal dynamic treatment for every individual patient. Providing such a schedule is likely to be one of the obstacles facing the implementation of the mechatronic valve. In this paper, an algorithm is proposed to help in developing such a schedule that dynamically change based on the patients' own intracranial pressure data and a novel figure of merit, thus providing the physician with an easy tool that facilitate the use of the mechatronic valve. The algorithm was implemented in M ATLAB and Simulink. Real ICP data for three hydrocephalus patients (before shunting) were used to test this algorithm and the resulted schedules along with the resulted intracranial pressure data have illustrated the effectiveness of the algorithm in providing schedule that maintain ICP within the normal limits.

Design of an intelligent and personalised shunting system for hydrocephalus.

Hydrocephalus is a neurological disease that manifests itself in an elevated fluid pressure within the brain, and if left untreated, may be fatal. It is currently treated using shunt implants, which consist of a mechanical valve and tubes that regulate the pressure of cerebrospinal fluid (CSF) by draining excess fluid into the abdomen. Hydrocephalus shunting systems are no longer expected simply to regulate the intracranial pressure (ICP), but also to offer the option of regaining independence of the shunt. Additionally, they could offer personalised valve management which is one of the main limitations of current shunts. This paper describes the design of a multi-agent system for an intelligent and personalised CSF management system. Patient feedback and intracranial pressure readings will play important roles in the process of CSF regulation and weaning, introduces an element of personalisation to the treatment. The new shunting system would deliver both reactive and goal-driven solutions for the treatment, at the same time the intelligent part of the system will be monitoring how well the shunt is performing. These tasks can be achieved by implementing an agent approach in designing this system. Such system would help us to understand more about the dynamics of hydrocephalus.