Transformer-Based Network for Accurate Classification of Lung Auscultation Sounds.

Respiratory diseases are a major cause of death worldwide, affecting a significant proportion of the population with lung function abnormalities that can lead to respiratory illnesses. Early detection and prevention are critical to effective management of these disorders. Deep learning algorithms offer a promising approach for analyzing complex medical data and aiding in early disease detection. While transformer-based models for sequence classification have proven effective for tasks like sentiment analysis, topic classification, etc., their potential for respiratory disease classification remains largely unexplored. This paper proposes a classifier utilizing the transformer-encoder block, which can capture complex patterns and dependencies in medical data. The proposed model is trained and evaluated on a large dataset from the International Conference on Biomedical Health Informatics 2017, achieving state-of-the-art results with a mean sensitivity of 70.53%, mean specificity of 84.10%, mean average score of 77.32%, and mean harmonic score of 76.10%. These results demonstrate the model's effectiveness in diagnosing respiratory diseases while taking up minimal computational resources.

Multivariate Parametric Study of Nailfold Capillary Images for Disease Detection.

Nailfold capillaroscopy is a tool which is non-invasive in nature and can be useful for diagnosis, research, therapeutic study and prognosis. Research shows that specific capillary morphology patterns are identified for diabetic subjects, hypertensive subjects and normal controls. In this study, we have proposed RATHEW approach of classifying these three classes of subjects. RATHEW approach employs a three step process for classifying nailfold images: one, identify six abnormality parameters from the image dataset; two, score these abnormality parameters based on the defined scoring rules; and three, combine them mathematically to segregate them into three classes. This technique can be further enhanced to grade the severity of disease and organ involvement. This can bring in a paradigm shift to the disease detection and therapeutic study mechanism.

The First Differentiated TB Care Model From India: Delays and Predictors of Losses in the Care Cascade.

To reduce TB deaths in resource-limited settings, a differentiated care strategy can be used to triage patients with high risk of severe illness (i.e., those with very severe undernutrition, respiratory insufficiency, or inability to stand without support) at diagnosis and refer them for comprehensive assessment and inpatient care. Globally, there are few examples of implementing this type of strategy in routine program settings. Beginning in April 2022, the Indian state of Tamil Nadu implemented a differentiated care strategy called Tamil Nadu-Kasanoi Erappila Thittam (TN-KET) for all adults aged 15 years and older with drug-susceptible TB notified by public facilities. Before evaluating the impact on TB deaths, we sought to understand the retention and delays in the care cascade as well as predictors of losses. During April-June 2022, 14,961 TB patients were notified and 11,599 (78%) were triaged. Of those triaged, 1,509 (13%) were at high risk of severe illness; of these, 1,128 (75%) were comprehensively assessed at a nodal inpatient care facility. Of 993 confirmed as severely ill, 909 (92%) were admitted, with 8% unfavorable admission outcomes (4% deaths). Median admission duration was 4 days. From diagnosis, the median delay in triaging and admission of severely ill patients was 1 day each. Likelihood of triaging decreased for people with extrapulmonary TB, those diagnosed in high-notification districts or teaching hospitals, and those transferred out of district. Predictors of not being comprehensively assessed included: aged 25-34 years, able to stand without support, and diagnosis at a primary or secondary-level facility. Inability to stand without support was a predictor of unfavorable admission outcomes. To conclude, the first quarter of implementation suggests that TN-KET was feasible to implement but could be improved by addressing predictors of losses in the care cascade and increasing admission duration.

--Eleven tips for operational researchers working with health programmes: our experience based on implementing differentiated tuberculosis care in south India.

Due to the workload and lack of a critical mass of trained operational researchers within their ranks, health systems and programmes may not be able to dedicate sufficient time to conducting operational research (OR). Hence, they may need the technical support of operational researchers from research/academic organisations. Additionally, there is a knowledge gap regarding implementing differentiated tuberculosis (TB) care in programme settings. In this 'how we did it' paper, we share our experience of implementing a differentiated TB care model along with an inbuilt OR component in Tamil Nadu, a southern state in India. This was a health system initiative through a collaboration of the State TB cell with the Indian Council of Medical Research institutes and the World Health Organisation country office in India. The learnings are in the form of eleven tips: four broad principles (OR on priority areas and make it a health system initiative, implement simple and holistic ideas, embed OR within routine programme settings, aim for long-term engagement), four related to strategic planning (big team of investigators, joint leadership, decentralised decision-making, working in advance) and three about implementation planning (conducting pilots, smart use of e-tools and operational research publications at frequent intervals). These may act as a guide for other Indian states, high TB burden countries that want to implement differentiated care, and for operational researchers in providing technical assistance for strengthening implementation and conducting OR in health systems and programmes (TB or other health programmes). Following these tips may increase the chances of i) an enriching engagement, ii) policy/practice change, and iii) sustainable implementation.

Electroencephalogram-Based Emotion Recognition Using a Particle Swarm Optimization-Derived Support Vector Machine Classifier.

We sort human emotions using Russell's circumplex model of emotion by classifying electroencephalogram (EEG) signals from 25 subjects into four discrete states, namely, happy, sad, angry, and relaxed. After acquiring signals, we use a standard database for emotion analysis using physiological EEG signals. Once raw signals are pre-processed in an EEGLAB, we perform feature extraction using Matrix Laboratory and apply discrete wavelet transform. Before classifying we optimize extracted features with particle swarm optimization. The acquired set of EEG signals are validated after finding average classification accuracy of 75.25%, average sensitivity of 76.8%, and average specificity of 91.06%.