Combining Deep Learning with Handcrafted Features for Cell Nuclei Segmentation.

Segmentation of cell nuclei in fluorescence microscopy images provides valuable information about the shape and size of the nuclei, its chromatin texture and DNA content. It has many applications such as cell tracking, counting and classification. In this work, we extended our recently proposed approach for nuclei segmentation based on deep learning, by adding to its input handcrafted features. Our handcrafted features introduce additional domain knowledge that nuclei are expected to have an approximately round shape. For round shapes the gradient vector of points at the border point to the center. To convey this information, we compute a map of gradient convergence to be used by the CNN as a new channel, in addition to the fluorescence microscopy image. We applied our method to a dataset of microscopy images of cells stained with DAPI. Our results show that with this approach we are able to decrease the number of missdetections and, therefore, increase the F1-Score when compared to our previously proposed approach. Moreover, the results show that faster convergence is obtained when handcrafted features are combined with deep learning.

COVID-19 pathways for brain and heart injury in comorbidity patients: A role of medical imaging and artificial intelligence-based COVID severity classification: A review.

Artificial intelligence (AI) has penetrated the field of medicine, particularly the field of radiology. Since its emergence, the highly virulent coronavirus disease 2019 (COVID-19) has infected over 10 million people, leading to over 500,000 deaths as of July 1st, 2020. Since the outbreak began, almost 28,000 articles about COVID-19 have been published (https://pubmed.ncbi.nlm.nih.gov); however, few have explored the role of imaging and artificial intelligence in COVID-19 patients-specifically, those with comorbidities. This paper begins by presenting the four pathways that can lead to heart and brain injuries following a COVID-19 infection. Our survey also offers insights into the role that imaging can play in the treatment of comorbid patients, based on probabilities derived from COVID-19 symptom statistics. Such symptoms include myocardial injury, hypoxia, plaque rupture, arrhythmias, venous thromboembolism, coronary thrombosis, encephalitis, ischemia, inflammation, and lung injury. At its core, this study considers the role of image-based AI, which can be used to characterize the tissues of a COVID-19 patient and classify the severity of their infection. Image-based AI is more important than ever as the pandemic surges and countries worldwide grapple with limited medical resources for detection and diagnosis.

Integration of cardiovascular risk assessment with COVID-19 using artificial intelligence.

Artificial Intelligence (AI), in general, refers to the machines (or computers) that mimic "cognitive" functions that we associate with our mind, such as "learning" and "solving problem". New biomarkers derived from medical imaging are being discovered and are then fused with non-imaging biomarkers (such as office, laboratory, physiological, genetic, epidemiological, and clinical-based biomarkers) in a big data framework, to develop AI systems. These systems can support risk prediction and monitoring. This perspective narrative shows the powerful methods of AI for tracking cardiovascular risks. We conclude that AI could potentially become an integral part of the COVID-19 disease management system. Countries, large and small, should join hands with the WHO in building biobanks for scientists around the world to build AI-based platforms for tracking the cardiovascular risk assessment during COVID-19 times and long-term follow-up of the survivors.

Wilson's disease: A new perspective review on its genetics, diagnosis and treatment.

Wilson's disease (WD) is an autosomal recessive disorder which is caused by poor excretion of copper in mammalian cells. In this review, various issues such as effective characterization of ATP7B genes, scope of gene network topology in genetic analysis, pattern recognition using different computing approaches and fusion possibilities in imaging and genetic dataset are discussed vividly. We categorized this study into three major sections: (A) WD genetics, (B) diagnosis guidelines and (3) treatment possibilities. We addressed the scope of advanced mathematical modelling paradigms for understanding common genetic sequences and dominating WD imaging biomarkers. We have also discussed current state-of-the-art software models for genetic sequencing. Further, we hypothesized that involvement of machine and deep learning techniques in the context of WD genetics and image processing for precise classification of WD. These computing procedures signify changing roles of various data transformation techniques with respect to supervised and unsupervised learning models.

Symtosis: A liver ultrasound tissue characterization and risk stratification in optimized deep learning paradigm.

Background and Objective Fatty Liver Disease (FLD) - a disease caused by deposition of fat in liver cells, is predecessor to terminal diseases such as liver cancer. The machine learning (ML) techniques applied for FLD detection and risk stratification using ultrasound (US) have limitations in computing tissue characterization features, thereby limiting the accuracy. Methods Under the class of Symtosis for FLD detection and risk stratification, this study presents a Deep Learning (DL)-based paradigm that computes nearly seven million weights per image when passed through a 22 layered neural network during the cross-validation (training and testing) paradigm. The DL architecture consists of cascaded layers of operations such as: convolution, pooling, rectified linear unit, dropout and a special block called inception model that provides speed and efficiency. All data analysis is performed in optimized tissue region, obtained by removing background information. We benchmark the DL system against the conventional ML protocols: support vector machine (SVM) and extreme learning machine (ELM). Results The liver US data consists of 63 patients (27 normal/36 abnormal). Using the K10 cross-validation protocol (90% training and 10% testing), the detection and risk stratification accuracies are: 82%, 92% and 100% for SVM, ELM and DL systems, respectively. The corresponding area under the curve is: 0.79, 0.92 and 1.0, respectively. We further validate our DL system using two class biometric facial data that yields an accuracy of 99%. Conclusion DL system shows a superior performance for liver detection and risk stratification compared to conventional machine learning systems: SVM and ELM.

Author Correction to: Extreme Learning Machine Framework for Risk Stratification of Fatty Liver Disease Using Ultrasound Tissue Characterization.

The original version of this article unfortunately contained a mistake. The family name of Rui Tato Marinho was incorrectly spelled as Marinhoe.

Extreme Learning Machine Framework for Risk Stratification of Fatty Liver Disease Using Ultrasound Tissue Characterization.

Fatty Liver Disease (FLD) is caused by the deposition of fat in liver cells and leads to deadly diseases such as liver cancer. Several FLD detection and characterization systems using machine learning (ML) based on Support Vector Machines (SVM) have been applied. These ML systems utilize large number of ultrasonic grayscale features, pooling strategy for selecting the best features and several combinations of training/testing. As result, they are computationally intensive, slow and do not guarantee high performance due to mismatch between grayscale features and classifier type. This study proposes a reliable and fast Extreme Learning Machine (ELM)-based tissue characterization system (a class of Symtosis) for risk stratification of ultrasound liver images. ELM is used to train single layer feed forward neural network (SLFFNN). The input-to-hidden layer weights are randomly generated reducing computational cost. The only weights to be trained are hidden-to-output layer which is done in a single pass (without any iteration) making ELM faster than conventional ML methods. Adapting four types of K-fold cross-validation (K = 2, 3, 5 and 10) protocols on three kinds of data sizes: S0-original, S4-four splits, S8-sixty four splits (a total of 12 cases) and 46 types of grayscale features, we stratify the FLD US images using ELM and benchmark against SVM. Using the US liver database of 63 patients (27 normal/36 abnormal), our results demonstrate superior performance of ELM compared to SVM, for all cross-validation protocols (K2, K3, K5 and K10) and all types of US data sets (S0, S4, and S8) in terms of sensitivity, specificity, accuracy and area under the curve (AUC). Using the K10 cross-validation protocol on S8 data set, ELM showed an accuracy of 96.75% compared to 89.01% for SVM, and correspondingly, the AUC: 0.97 and 0.91, respectively. Further experiments also showed the mean reliability of 99% for ELM classifier, along with the mean speed improvement of 40% using ELM against SVM. We validated the symtosis system using two class biometric facial public data demonstrating an accuracy of 100%.

An ultrasonographic risk score for detecting symptomatic carotid atherosclerotic plaques.

this paper proposes a risk score computed from ultrasound data that correlates to plaque activity. It has the twofold purpose of detecting symptomatic plaques and estimating the likelihood of the asymptomatic lesion to become symptomatic. The proposed ultrasonographic activity index (UAI) relies on the plaque active profile, which is a combination of the most discriminate ultrasound parameter associated with symptoms. These features are extracted by the automatic algorithm and also by the physician from the ultrasound images and from some transformations on it, such as monogenic decomposition, which is a novelty in this clinical problem. This information is used to compute a risk score from the conditional probabilities of either symptomatic or asymptomatic groups. Symptom detection performance is evaluated on a transversal dataset of 146 plaques, where UAI obtained 83.5% accuracy, 84.1% sensitivity, and 83.7% specificity. Performance is also assessed on a longitudinal study of 112 plaques, where UAI shows a significant improvement over the gold standard degree of stenosis, demonstrating higher power at predicting which asymptomatic plaques developed symptoms in an average follow-up of ten months. Results suggest that this score could have a positive impact on early stroke prevention and treatment planning.

Hypnogram and sleep parameter computation from activity and cardiovascular data.

The automatic computation of the hypnogram and sleep Parameters, from the data acquired with portable sensors, is a challenging problem with important clinical applications. In this paper, the hypnogram, the sleep efficiency (SE), rapid eye movement (REM), and nonREM (NREM) sleep percentages are automatically estimated from physiological (ECG and respiration) and behavioral (Actigraphy) nocturnal data. Two methods are described; the first deals with the problem of the hypnogram estimation and the second is specifically designed to compute the sleep parameters, outperforming the traditional estimation approach based on the hypnogram. Using an extended set of features the first method achieves an accuracy of 72.8%, 77.4%, and 80.3% in the detection of wakefulness, REM, and NREM states, respectively, and the second an estimation error of 4.3%, 9.8%, and 5.4% for the SE, REM, and NREM percentages, respectively.

An ultrasound-based computer-aided diagnosis tool for steatosis detection.

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. In this paper, a new computer-aided diagnosis (CAD) system for steatosis classification, in a local and global basis, is presented. A Bayes factor is computed from objective ultrasound textural features extracted from the liver parenchyma. The goal is to develop a CAD screening tool to help in the steatosis detection. Results showed an accuracy of 93.33%, with a sensitivity of 94.59% and specificity of 92.11%, using the Bayes classifier. The proposed CAD system is a suitable graphical display for steatosis classification.

Asymptomatic carotid disease--a new tool for assessing neurological risk.

Active carotid plaques are associated with atheroembolism and neurological events; its identification is crucial for stroke prevention. High-definition ultrasound (HDU) can be used to recognize plaque structure in carotid bifurcation stenosis associated with plaque vulnerability and occurrence of brain ischemic events. A new computer-assisted HDU method to study the echomorphology of the carotid plaque and to determine a risk score for developing appropriate symptoms is proposed in this study. Plaque echomorphology characteristics such as presence of ulceration at the plaque surface, juxta-luminal location of echolucent areas, echoheterogeneity were obtained from B-mode ultrasound scans using several image processing algorithms and were combined with measurement of severity of stenosis to obtain a clinical score--enhanced activity index (EAI)--which was correlated with the presence or absence of ipsilateral appropriate ischemic symptoms. An optimal cutoff value of EAI was determined to obtain the best separation between symptomatic (active) from asymptomatic (inactive) plaques and its diagnostic yield was compared to other 2 reference methods by means of receiver-operating characteristic (ROC) analysis. Classification performance was evaluated by leave-one-patient-out cross-validation applied to a cohort of 146 carotid plaques from 99 patients. The proposed method was benchmarked against (a) degree of stenosis criteria and (b) earlier proposed activity index (AI) and demonstrated that EAI yielded the highest accuracy up to an accuracy of 77% to predict asymptomatic plaques that developed symptoms in a prospective cross-sectional study. Enhanced activity index is a noninvasive, easy to obtain parameter, which provided accurate estimation of neurological risk of carotid plaques.

Sleep and wakefulness state detection in nocturnal actigraphy based on movement information.

Wrist actigraphy (ACT) is a low-cost and well-established technique for long-term monitoring of human activity. It has a special relevance in sleep studies, where its noninvasive nature makes it a valuable tool for behavioral characterization and for the detection and diagnosis of some sleep disorders. The traditional sleep/wakefulness state estimation algorithms from the nocturnal ACT data are unbalanced from a sensitivity and specificity points of view since they tend to overestimate sleep state, with severe consequences from a diagnosis point of view. They usually maximize the overall accuracy that does not take into account the highly unbalanced state distribution. In this paper, a method is proposed to appropriately deal with this unbalanced problem, achieving similar sensitivity and specificity scores in the state estimation process. The proposed method combines two linear discriminant classifiers, trained with two different criteria involving movement detection to generate a first state estimate. This result is then refined by a Hidden Markov Model-based algorithm. The global accuracy, the sensitivity, and the specificity of the method are 77.8%, 75.6%, and 81.6%, respectively, performing better than the tested algorithms. If the performance is assessed only for movement periods, this improvement is even higher.

Photobleaching/photoblinking differential equation model for fluorescence microscopy imaging.

Fluorescence images present low signal-to-noise ratio (SNR), are corrupted by a type of multiplicative noise with Poisson distribution, and are affected by a time intensity decay due to photoblinking and photobleaching (PBPB) effects. The noise and the PBPB effects together make long-term biological observation very difficult. Here, a theoretical model based on the underlying quantum mechanic physics theory of the observation process associated with this type of image is presented and the common empirical weighted sum of two decaying exponentials is derived from the model. Improvement in the SNR obtained in denoising when the proposed method is used is particularly important in the last images of the sequence where temporal correlation is used to recover information that is sometimes faded and therefore useless from a visual inspection point of view. The proposed PBPB model is included in a Bayesian denoising algorithm previously proposed by the authors. Experiments with synthetic and real data are presented to validate the PBPB model and to illustrate the effectiveness of the model in denoising and reconstruction results.

Classification and staging of chronic liver disease from multimodal data.

Chronic liver disease (CLD) is most of the time an asymptomatic, progressive, and ultimately potentially fatal disease. In this study, an automatic hierarchical procedure to stage CLD using ultrasound images, laboratory tests, and clinical records are described. The first stage of the proposed method, called clinical based classifier (CBC), discriminates healthy from pathologic conditions. When nonhealthy conditions are detected, the method refines the results in three exclusive pathologies in a hierarchical basis: 1) chronic hepatitis; 2) compensated cirrhosis; and 3) decompensated cirrhosis. The features used as well as the classifiers (Bayes, Parzen, support vector machine, and k -nearest neighbor) are optimally selected for each stage. A large multimodal feature database was specifically built for this study containing 30 chronic hepatitis cases, 34 compensated cirrhosis cases, and 36 decompensated cirrhosis cases, all validated after histopathologic analysis by liver biopsy. The CBC classification scheme outperformed the nonhierachical one against all scheme, achieving an overall accuracy of 98.67% for the normal detector, 87.45% for the chronic hepatitis detector, and 95.71% for the cirrhosis detector.

A CAD system for atherosclerotic plaque assessment.

Recently, several atherosclerotic plaque characterization methods were proposed based on plaque morphology assessed through 2D ultrasound. It is of extreme importance to establish an objective quantification measure which allows the physicians to determine the risk of plaque rupture, and thus, of brain stroke. Having these, sometimes complex, measures easily and quickly assessed might prove invaluable for the physician an patient alike. This paper is a first attempt to incorporate such scores in a user-friendly software platform for Computer-aided Diagnosis. This tool provides a way to objectively and interactively characterize the atherosclerotic plaque, to store relevant patient data and to use several processing tools to outline the plaque and compute different echogenicity measures. Combinations of these features are used to provide two objective measure with clinical significance, known as activity index and enhanced activity index.

Phase-locking factor in a motor imagery Brain-Computer Interface.

A Brain-Computer Interface (BCI) attempts to create a direct channel of communication between the brain and a computer. This is especially important for patients that are "locked in", as they have limited motor function and thus require an alternative means of communication. In this scope, a BCI can be controlled through the imagination of motor tasks, i.e. Motor Imagery. This thinking of actions produce changes on the ongoing Electroencephalogram (EEG), such as the so called Event-Related Desynchronization (ERD), that can be detected and measured. Traditionally, ERD is measured through the estimation of EEG signal power in specific frequency bands. In this work, a new method based on the phase information from the EEG channels, through the Phase-Locking Factor (PLF), is proposed. Both feature types were tested in real data obtained from 6 voluntary subjects, who performed 7 motor tasks in an EEG session. The features were classified using Support Vector Machine (SVM) classifiers organized in a hierarchical structure. The results show that the PLF features are better, with an average accuracy of ≈ 86%, against an accuracy of ≈ 70% for the band power features. Although more research is still needed, the PLF measure shows promising results for use in a BCI system.

Global and local detection of liver steatosis from ultrasound.

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.

Optimal sampling and estimation in PASL perfusion imaging.

Pulsed arterial spin labeling (PASL) techniques potentially allow the absolute, noninvasive quantification of brain perfusion using MRI. This can be achieved by fitting a kinetic model to the data acquired at a number of sampling times. However, the intrinsically low signal-to-noise ratio of PASL measurements usually requires substantial signal averaging, which may result in undesirably long scanning times. A judicious choice of the sampling points is, therefore, crucial in order to minimize scanning time, while optimizing estimation accuracy. On the other hand, a priori information regarding the model parameters may improve estimation performance. Here, we propose a Bayesian framework to determine an optimal sampling strategy and estimation method for the measurement of brain perfusion and arterial transit time (ATT). A Bayesian Fisher information criterion is used to determine the optimal sampling points and a MAP criterion is employed for the estimation of the model parameters, both taking into account the uncertainty in the model parameters as well as the amount of noise in the data. By Monte Carlo simulations, we show that using optimal compared to uniform sampling strategies, as well as the Bayesian estimator relative to a standard least squares approach, improves the accuracy of perfusion and ATT measurements. Moreover, we also demonstrate the applicability of the proposed approach to real data, with the advantage of reduced intersubject variability relative to conventional sampling and estimation approaches.

The usefulness of ultrasound in the classification of chronic liver disease.

Chronic Liver Disease is a progressive, most of the time asymptomatic, and potentially fatal disease. In this paper, a semi-automatic procedure to stage this disease is proposed based on ultrasound liver images, clinical and laboratorial data. In the core of the algorithm two classifiers are used: a k nearest neighbor and a Support Vector Machine, with different kernels. The classifiers were trained with the proposed multi-modal feature set and the results obtained were compared with the laboratorial and clinical feature set. The results showed that using ultrasound based features, in association with laboratorial and clinical features, improve the classification accuracy. The support vector machine, polynomial kernel, outperformed the others classifiers in every class studied. For the Normal class we achieved 100% accuracy, for the chronic hepatitis with cirrhosis 73.08%, for compensated cirrhosis 59.26% and for decompensated cirrhosis 91.67%.

Statistical characterization of actigraphy data during sleep and wakefulness states.

Human activity can be measured with actimetry sensors used by the subjects in several locations such as the wrists or legs. Actigraphy data is used in different contexts such as sports training or tele-medicine monitoring. In the diagnosis of sleep disorders, the actimetry sensor, which is basically a 3D axis accelerometer, is used by the patient in the non dominant wrist typically during an entire week. In this paper the actigraphy data is described by a weighted mixture of two distributions where the weight evolves along the day according to the patient circadian cycle. Thus, one of the distributions is mainly associated with the wakefulness state while the other is associated with the sleep state. Actigraphy data, acquired from 20 healthy patients and manually segmented by trained technicians, is used to characterize the acceleration magnitude during sleep and wakefulness states. Several mixture combinations are tested and statistically validated with conformity measures. It is shown that both distributions can co-exist at a certain time with varying importance along the circadian cycle.