AI-Driven Precision: Integrating Multiscale Transformer Fusion Network and Bio-Inspired Optimization for CKD Classification

This research paper presents an innovative approach to Chronic Kidney Disease (CKD) classification by optimizing a Multiscale Transformer Fusion Network (MS-TFN) using Electric Eel Foraging Optimization (EEFO). A dataset of 1000 instances with 25 features, including patient demographics and health indicators, was preprocessed using categorical encoding and numerical normalization before being split into training and test sets. The MS-TFN integrates transformer blocks for capturing long-range dependencies with depthwise separable convolutional blocks for efficient local feature extraction, allowing for the creation of robust data representations that are processed through dense layers for accurate classification. The EEFO, a bio-inspired optimization algorithm, fine-tunes the learning rate and batch size by balancing exploration and exploitation. This synergy achieved a classification accuracy of 99.9%, demonstrating significant potential for clinical application. However, the study's limitation lies in the generalizability of the results, as the dataset is limited in size and population scope, requiring further validation with more diverse datasets. This work introduces a novel combination of transformer and convolutional blocks in the MS-TFN architecture alongside bio-inspired EEFO, offering a promising tool for early CKD diagnosis and personalized treatment plans in clinical settings.