Integration of Localized, Contextual, and Hierarchical Features in Deep Learning for Improved Skin Lesion Classification.

Skin lesion classification is vital for the early detection and diagnosis of skin diseases, facilitating timely intervention and treatment. However, existing classification methods face challenges in managing complex information and long-range dependencies in dermoscopic images. Therefore, this research aims to enhance the feature representation by incorporating local, global, and hierarchical features to improve the performance of skin lesion classification. We introduce a novel dual-track deep learning (DL) model in this research for skin lesion classification. The first track utilizes a modified Densenet-169 architecture that incorporates a Coordinate Attention Module (CoAM). The second track employs a customized convolutional neural network (CNN) comprising a Feature Pyramid Network (FPN) and Global Context Network (GCN) to capture multiscale features and global contextual information. The local features from the first track and the global features from second track are used for precise localization and modeling of the long-range dependencies. By leveraging these architectural advancements within the DenseNet framework, the proposed neural network achieved better performance compared to previous approaches. The network was trained and validated using the HAM10000 dataset, achieving a classification accuracy of 93.2%.

A hybrid model for the classification of Autism Spectrum Disorder using Mu rhythm in EEG.

BACKGROUND: Autism Spectrum Disorder (ASD) is a condition with social interaction, communication, and behavioral difficulties. Diagnostic methods mostly rely on subjective evaluations and can lack objectivity. In this research Machine learning (ML) and deep learning (DL) techniques are used to enhance ASD classification. OBJECTIVE: This study focuses on improving ASD and TD classification accuracy with a minimal number of EEG channels. ML and DL models are used with EEG data, including Mu Rhythm from the Sensory Motor Cortex (SMC) for classification. METHODS: Non-linear features in time and frequency domains are extracted and ML models are applied for classification. The EEG 1D data is transformed into images using Independent Component Analysis-Second Order Blind Identification (ICA-SOBI), Spectrogram, and Continuous Wavelet Transform (CWT). RESULTS: Stacking Classifier employed with non-linear features yields precision, recall, F1-score, and accuracy rates of 78%, 79%, 78%, and 78% respectively. Including entropy and fuzzy entropy features further improves accuracy to 81.4%. In addition, DL models, employing SOBI, CWT, and spectrogram plots, achieve precision, recall, F1-score, and accuracy of 75%, 75%, 74%, and 75% respectively. The hybrid model, which combined deep learning features from spectrogram and CWT with machine learning, exhibits prominent improvement, attained precision, recall, F1-score, and accuracy of 94%, 94%, 94%, and 94% respectively. Incorporating entropy and fuzzy entropy features further improved the accuracy to 96.9%. CONCLUSIONS: This study underscores the potential of ML and DL techniques in improving the classification of ASD and TD individuals, particularly when utilizing a minimal set of EEG channels.

A dual-track feature fusion model utilizing Group Shuffle Residual DeformNet and swin transformer for the classification of grape leaf diseases.

Grape cultivation is important globally, contributing to the agricultural economy and providing diverse grape-based products. However, the susceptibility of grapes to disease poses a significant threat to yield and quality. Traditional disease identification methods demand expert knowledge, which limits scalability and efficiency. To address these limitations our research aims to design an automated deep learning approach for grape leaf disease detection. This research introduces a novel dual-track network for classifying grape leaf diseases, employing a combination of the Swin Transformer and Group Shuffle Residual DeformNet (GSRDN) tracks. The Swin Transformer track exploits shifted window techniques to construct hierarchical feature maps, enhancing global feature extraction. Simultaneously, the GSRDN track combines Group Shuffle Depthwise Residual block and Deformable Convolution block to extract local features with reduced computational complexity. The features from both tracks are concatenated and processed through Triplet Attention for cross-dimensional interaction. The proposed model achieved an accuracy of 98.6%, the precision, recall, and F1-score are recorded as 98.7%, 98.59%, and 98.64%, respectively as validated on a dataset containing grape leaf disease information from the PlantVillage dataset, demonstrating its potential for efficient grape disease classification.

CT-based severity assessment for COVID-19 using weakly supervised non-local CNN.

Evaluating patient criticality is the foremost step in administering appropriate COVID-19 treatment protocols. Learning an Artificial Intelligence (AI) model from clinical data for automatic risk-stratification enables accelerated response to patients displaying critical indicators. Chest CT manifestations including ground-glass opacities and consolidations are a reliable indicator for prognostic studies and show variability with patient condition. To this end, we propose a novel attention framework to estimate COVID-19 severity as a regression score from a weakly annotated CT scan dataset. It takes a non-locality approach that correlates features across different parts and spatial scales of the 3D scan. An explicit guidance mechanism from limited infection labeling drives attention refinement and feature modulation. The resulting encoded representation is further enriched through cross-channel attention. The attention model also infuses global contextual awareness into the deep voxel features by querying the base CT scan to mine relevant features. Consequently, it learns to effectively localize its focus region and chisel out the infection precisely. Experimental validation on the MosMed dataset shows that the proposed architecture has significant potential in augmenting existing methods as it achieved a 0.84 R-squared score and 0.133 mean absolute difference.

Contour-enhanced attention CNN for CT-based COVID-19 segmentation.

Accurate detection of COVID-19 is one of the challenging research topics in today's healthcare sector to control the coronavirus pandemic. Automatic data-powered insights for COVID-19 localization from medical imaging modality like chest CT scan tremendously augment clinical care assistance. In this research, a Contour-aware Attention Decoder CNN has been proposed to precisely segment COVID-19 infected tissues in a very effective way. It introduces a novel attention scheme to extract boundary, shape cues from CT contours and leverage these features in refining the infected areas. For every decoded pixel, the attention module harvests contextual information in its spatial neighborhood from the contour feature maps. As a result of incorporating such rich structural details into decoding via dense attention, the CNN is able to capture even intricate morphological details. The decoder is also augmented with a Cross Context Attention Fusion Upsampling to robustly reconstruct deep semantic features back to high-resolution segmentation map. It employs a novel pixel-precise attention model that draws relevant encoder features to aid in effective upsampling. The proposed CNN was evaluated on 3D scans from MosMedData and Jun Ma benchmarked datasets. It achieved state-of-the-art performance with a high dice similarity coefficient of 85.43% and a recall of 88.10%.

Deep learning-based school attendance prediction for autistic students.

Autism Spectrum Disorder is a neurodevelopmental disorder characterized by deficits in social communication and interaction as well as the presence of repetitive, restricted patterns of behavior, interests, or activities. Many autistic students experience difficulty with daily functioning at school and home. Given these difficulties, regular school attendance is a primary source for autistic students to receive an appropriate range of needed educational and therapeutic interventions. Moreover, school absenteeism (SA) is associated with negative consequences such as school drop-out. Therefore, early SA prediction would help school districts to intervene properly to ameliorate this issue. Due to its heterogeneity, autistic students show within-group differences concerning their SA. A comprehensive statistical analysis performed by the authors shows that the individual and demographic characteristics of the targeted population are not predictive factors of SA. So, we used the students' recent previous attendance to predict their future attendance. We introduce a deep learning-based framework for predicting short-and long-term SA of autistic students using the Long Short-Term Memory (LSTM) and Multilayer Perceptron (MLP) algorithms. The adopted algorithms outperform other machine learning algorithms. In detail, LSTM increased the accuracy and recall of short-term SA prediction by 20% and 13%, while the same scores of long-term SA prediction increased by 5% using MLP.

Automatic cell counting from stimulated Raman imaging using deep learning.

In this paper, we propose an automatic cell counting framework for stimulated Raman scattering (SRS) images, which can assist tumor tissue characteristic analysis, cancer diagnosis, and surgery planning processes. SRS microscopy has promoted tumor diagnosis and surgery by mapping lipids and proteins from fresh specimens and conducting a fast disclose of fundamental diagnostic hallmarks of tumors with a high resolution. However, cell counting from label-free SRS images has been challenging due to the limited contrast of cells and tissue, along with the heterogeneity of tissue morphology and biochemical compositions. To this end, a deep learning-based cell counting scheme is proposed by modifying and applying U-Net, an effective medical image semantic segmentation model that uses a small number of training samples. The distance transform and watershed segmentation algorithms are also implemented to yield the cell instance segmentation and cell counting results. By performing cell counting on SRS images of real human brain tumor specimens, promising cell counting results are obtained with > 98% of area under the curve (AUC) and R = 0.97 in terms of cell counting correlation between SRS and histological images with hematoxylin and eosin (H&E) staining. The proposed cell counting scheme illustrates the possibility and potential of performing cell counting automatically in near real time and encourages the study of applying deep learning techniques in biomedical and pathological image analyses.

Ischemic Lesion Segmentation using Ensemble of Multi-Scale Region Aligned CNN.

The first and foremost step in the diagnosis of ischemic stroke is the delineation of the lesion from radiological images for effective treatment planning. Manual delineation of the lesion by radiological experts is generally laborious and time-consuming. Sometimes, it is prone to intra-observer and inter-observer variability. State of the art deep architectures based on Fully Convolutional Networks (FCN) and cascaded CNNs have shown good results in automated lesion segmentation. This work proposes a series of enhancements over the learning paradigm in the existing methods, by focusing on learning meticulous feature representations through the CNN layers for accurate ischemic lesion segmentation from multimodal MRI. Multiple levels of losses, integration of features from multiple scales, an ensemble of prediction maps from sub-networks are employed to enable the CNN to correlate between features seen from different receptive fields. To allow for progressive refinement of features from block to block, a custom dropout module has been proposed that suppresses noisy features. Multi-branch residual connections and attention mechanisms were also included in the CNN blocks to enable the integration of information from multiple receptive fields and selectively weigh significant features. Also, to tackle data imbalance both at voxel and sample level, patch-based modeling and separation of concerns into classification & segmentation functional branches are proposed. By incorporating the above mentioned architectural enhancements, the proposed deep architecture was able to achieve better segmentation performance against the existing models. The proposed approach was evaluated on the ISLES 2015 SISS dataset, and it achieved a mean dice coefficient of 0.775. By combining sample classification and lesion segmentation into a fully automated framework, the proposed approach has yielded better results compared to most of the existing works.

Prediction of Mental Illness in Heart Disease Patients: Association of Comorbidities, Dietary Supplements, and Antibiotics as Risk Factors.

Comorbidities, dietary supplement use, and prescription drug use may negatively (or positively) affect mental health in cardiovascular patients. Although the significance of mental illnesses, such as depression, anxiety, and schizophrenia, on cardiovascular disease is well documented, mental illnesses resulting from heart disease are not well studied. In this paper, we introduce the risk factors of mental illnesses as an exploratory study and develop a prediction framework for mental illness that uses comorbidities, dietary supplements, and drug usage in heart disease patients. Particularly, the data used in this study consist of the records of 68,647 patients with heart disease, including the patient's mental illness information and the patient's intake of dietary supplements, antibiotics, and comorbidities. Patients in age groups <61, gender differences, and drug intakes, such as Azithromycin, Clarithromycin, Vitamin B6, and Coenzyme Q10, were associated with mental illness. For predictive modeling, we consider applying various state-of-the-art machine learning techniques with tuned parameters and finally obtain the following: Depression: 78.01% accuracy, 79.13% sensitivity, 72.65% specificity, and 86.26% Area Under the Curve (AUC). Anxiety: 82.93% accuracy, 82.86% sensitivity, 83.35% specificity, and 88.45% AUC. Schizophrenia: 87.59% accuracy, 87.70% sensitivity, 85.14% specificity, and 92.73% AUC. Disease: 86.63% accuracy, 95.50% sensitivity, 77.76% specificity, and 91.59% AUC. From the results, we conclude that using heart disease information, comorbidities, dietary supplement use, and antibiotics enables us to accurately predict the mental health outcome.

Sweat and saliva cortisol response to stress and nutrition factors.

Cortisol is a biomarker for stress monitoring; however, the biomedical and clinical relevance is still controversial due to the complexity of cortisol secretion mechanisms and their circadian cycles as well as environmental factors that affect physiological cortisol level, which include individual mood and dietary intake. To further investigate this multifaceted relationship, a human pilot study examined cortisol concentration in sweat and saliva samples collected from 48 college-aged participants during aerobic exercise sessions along with mental distress and nutrition surveys. Enzyme-linked immunosorbent assays determined highly significant differences between apocrine-dominant sweat (AP), saliva before exercise (SBE), and saliva after exercise (SAE) cortisol concentration (AP-SBE: p = 0.0017, AP-SAE: p = 0.0102). A significantly greater AP cortisol concentration was detected in males compared to females (p = 0.0559), and significant SAE cortisol concentration differences were also recorded between recreational athletes and non-athletes (p = 0.044). However, Kessler 10 Psychological Distress Scale (K10) scores, an examination administered to deduce overall wellness, provided no significant differences between males and females or athletes and non-athletes in distress levels, which statistically signifies a direct relationship to cortisol was not present. For further analysis, dietary intake from all participants was considered to investigate whether a multiplexed association was prevalent between nutrition, mood, and cortisol release. Significant positive correlations between AP cortisol, SAE cortisol, K10 scores, and fat intake among female participants and athletes were discovered. The various machine learning algorithms utilized the extensive connections between dietary intake, overall well-being, sex factors, athletic activity, and cortisol concentrations in various biofluids to predict K10 scores. Indeed, the understanding of physiochemical stress response and the associations between studied factors can advance algorithm developments for cortisol biosensing systems to mitigate stress-based illnesses and improve an individual's quality of life.

Investigating electroencephalography signals of autism spectrum disorder (ASD) using Higuchi Fractal Dimension.

Autism spectrum disorder (ASD) is a neurodevelopmental disorder with a deficit of social relationships, interaction, sense of imagination, and constrained interests. Early diagnosis of ASD will aid in devising appropriate training procedures and placing those children in the normal stream. The objective of this research is to analyze the brain response for auditory/visual stimuli in Typically Developing (TD) and children with autism through electroencephalography (EEG). Brain dynamics in the EEG signal can be analyzed well with the help of nonlinear feature primitives. Recent research reveals that, application of fractal-based techniques proves to be effective to estimate of degree of nonlinearity in a signal. This research attempts to analyze the effect of brain dynamics with Higuchi Fractal Dimension (HFD). Also, the performance of the fractal based techniques depends on the selection of proper hyper-parameters involved in it. One of the key parameters involved in computation of HFD is the time interval parameter 'k'. Most of the researches arbitrarily fixes the value of 'k' in the range of all channels. This research proposes an algorithm to estimate the optimal value of the time parameter for each channel. Sub-band analysis was also carried out for the responding channels. Statistical analysis on the experimental reveals that a difference of 30% was observed between autistic and Typically Developing children.

Network Optimization of Functional Connectivity Within Default Mode Network Regions to Detect Cognitive Decline.

The rapid aging of the world's population is causing an increase in the prevalence of cognitive decline and degenerative brain disease in the elderly. Current diagnoses of amnestic and nonamnestic mild cognitive impairment, which may represent early stage Alzheimer's disease or related degenerative conditions, are based on clinical grounds. The recent emergence of advanced network analyses of functional magnetic resonance imaging (fMRI) data taken at cognitive rest has provided insight that declining functional connectivity of the default mode network (DMN) may be correlated with neurological disorders, and particularly prodromal Alzheimer's disease. The goal of this paper is to develop a network analysis technique using fMRI data to characterize transition stages from healthy brain aging to cognitive decline. Previous studies primarily focused on inter-nodal connectivity of the DMN and often assume functional homogeneity within each DMN region. In this paper, we develop a technique that focuses on identifying critical intra-nodal DMN connectivity by incorporating sparsity into connectivity modeling of the k -cardinality tree (KCT) problem. Most biological networks are efficient and formed by sparse connections, and the KCT can potentially reveal sparse connectivity patterns that are biologically informative. The KCT problem is NP-hard, and existing solution approaches are mostly heuristic. Mathematical formulations of the KCT problem in the literature are not compact and do not provide good solution bounds. This paper presents new KCT formulations and a fast heuristic approach to efficiently solve the KCT models for large DMN regions. The results in this paper demonstrate that traditional fMRI group analysis on DMN regions cannot detect any statistically significant connectivity differences between normal aging and cognitively impaired subjects in DMN regions, and the proposed KCT approaches are more sensitive than the state-of-the-art regional homogeneity approach in detecting significant differences in both left and right medial temporal regions of the DMN.

Altered visual contrast gain control is sensitive for idiopathic generalized epilepsies.

OBJECTIVE: Visual hyperexcitability in the form of abnormal contrast gain control has been shown in photosensitive epilepsy and idiopathic generalized epilepsies. We assessed the accuracy and reliability of measures of visual contrast gain control in discerning individuals with idiopathic generalized epilepsies from healthy controls. METHODS: Twenty-four adult patients with idiopathic generalized epilepsy and 32 neurotypical control subjects from two study sites participated in a prospective, cross-sectional study. We recorded steady-state visual evoked potentials to a wide range of contrasts of a flickering grating stimulus. The resultant response magnitude vs. contrast curves were fitted to a standard model of contrast response function, and the model parameters were used as input features to a linear classifier to separate patients from controls. Additionally we compared the relative contribution of model parameters towards the classification using a sparse feature-selection approach. RESULTS: Classification accuracy was 80% or better. Sensitivity and specificity both were 80-85%. Cross validation confirmed robust classifier performance generalizable across the data from the two samples. Patients' relative lack of gain control at high contrasts was the most important information distinguishing patients from controls. CONCLUSIONS: Individuals with idiopathic generalized epilepsy were distinguishable from the neurotypical with a high degree of accuracy and reliability by a reduction in gain control at high contrasts. SIGNIFICANCE: Gain control is an essential neural operation that regulates neuronal sensitivity to stimuli and may represent a novel biomarker of hyperexcitability.