Research

Heart failure presents considerable mortality risks, highlighting the need for reliable predictors to support effective patient prognosis. This study explores the predictive power of specific clinical indicators, including ejection fraction, kidney function, smoking status, and follow-up time, for mortality in heart failure patients. Using a balanced training dataset, we conducted feature selection based on validation set accuracy, identifying these four indicators as the most influential subset. Models were subsequently tested on an unseen test set, with Random Forest achieving the highest accuracy at 87%, followed by Support Vector Machine (SVM) at 83.3%, and Logistic Regression at 80%. When all 12 initial indicators were applied, accuracy across all models converged to 83.3%. These findings emphasize the role of targeted feature selection in improving model performance, suggesting that focusing on key predictors can streamline clinical assessments while enhancing prognostic accuracy. This study underscores the potential of refined indica- tor selection to optimize resource allocation and improve patient outcomes in heart failure management.

Paper

This study explores the application of EfficientNet combined with Random Forest for classifying chest X-ray images to detect and categorize pneumonia. Initially, the model was em- ployed to classify X-ray images as either normal or pneumonia- affected, achieving an accuracy of 98.4%. Subsequently, pneu- monia cases were further categorized into three distinct types: COVID-19-induced, bacterial, and viral pneumonia, yielding a classification accuracy of 83%. The feature extraction capability of EfficientNet, a state-of-the-art convolutional neural network pre-trained on ImageNet, was leveraged to distill meaningful patterns from the medical images, while Random Forest (known for handling complex, non-linear relationships), served as the classifier for final decision-making process. Notably, the model effectively differentiated between bacterial and viral pneumonia, achieving an accuracy of 86.91%, despite the overlapping features between these two types. This dual-stage approach (first iden- tifying the presence of pneumonia, followed by specific subtype classification) demonstrates the potential for AI-driven diagnostic tools to assist in more accurate and detailed pneumonia diagnosis.

Paper

Face recognition by familiarity or recollection is a task people perform routinely in their daily lives. In the process of automating human experiences, existing studies have applied traditional machine learning applications and deep learning techniques on enough datasets (samples >= 1000) for human faces classification. However, the application of deep learning on electroencephalography (EEG) for new and familiar faces classification with limited data (samples < 100) has not been studied. We devised a face familiarity judgment EEG experiment and recruited eleven (11) participants for our study. We represented each trial by a visualization technique upon the generated EEG. The average power bands (theta, alpha, lower beta, higher beta, and gamma) from each channel at every 125ms window were computed and combined to form an image. We applied “ConvoForest,” a combination of convolution neural network (CNN) and random forest for classification. In comparison with conventional CNN where the dense layer was present, “ConvoForest” performed better with an average subject-dependent classification accuracy of 79.0% and an F1 score of 0.8.

Paper

Recently, there has been an exponential expansion in research focusing on AI-based music generation. Our in-depth analysis of the arXiv dataset revealed a growing number of publications on this subject: over 273 AI music papers in the past two years, with 102 explicitly tackling AI music generation. However, one area that seems underrepresented is Indian traditional music. This study presents the application of artificial intelligence (AI) in creating Indian classical music, focusing on Raga-based music generation. We outline the two-stage music creation process, including the creative and technical aspects, and explore how AI can be integrated into these stages. We trained the models using the LSTM (Long Short-term Memory network) and Transformer models on the Dunya dataset, which includes almost 250 ragas played across 12 instruments. Further, the study proposes a new Raga Multi-Track Music Model (RMMM) model to generate multi-layered Raga-based music with enhanced authenticity and emotional resonance. Despite potential challenges, this research opens an exciting journey in AI-generated Indian classical music.

Paper