Srinivasan, RRSrinivasanKorah, ReebaRavichandran, M2026-02-092026-02-092025https://gnanaganga.alliance.edu.in/handle/123456789/9267Neikuri, a traditional diagnostic technique in Siddha medicine, involves visual interpretation of oil dispersion patterns in urine samples for prognostic evaluation. While medically significant, the method remains largely subjective and practitionerdependent, lacking standardization and reproducibility. This doctoral research aims to modernize Neikuri analysis by developing an AI-enabled diagnostic system that integrates image processing with state-of-the-art deep learning algorithms. The image acquisition hardware developed for this study was a critical component in ensuring that images captured from urine-oil samples were consistent and reliable for AI training. It featured a controlled lighting chamber, an adjustable yet fixed camera mount and a standardized sample tray designed to maintain uniform distance, angle and positioning across sessions. This carefully calibrated physical parameter eliminated shadows, glare and perspective distortion—challenges that previously compromised image usability. This device was not only essential for this study but also represents a valuable tool for the broader research community interested in digitizing traditional diagnostic methods. The system was formally patented, offering a standardized and replicable solution for Neikuri image capture. The annotated dataset, curated in collaboration with expert Siddha practitioners from Government College of Siddha, was used to train and evaluate multiple Convolutional Neural Network (CNN) architectures (DenseNet, ResNet, Inception, VGG19 and EfficientNet). DenseNet model achieved the highest classification accuracy of 93.33%, demonstrating robust generalization and training stability. This research also explores parallels with Ayurveda's Taila Bindu Pariksha, highlighting interdisciplinary relevance across Indian traditional medicine systems. By bridging the ancient diagnostic wisdom with modern AI methodologies, this study aims to contribute a scalable, objective and reproducible framework that supports the integration of traditional practices into contemporary healthcare systems. The clinical implications of a non-invasive, urine-based diagnostic test capable of offering valuable prognostic insights are substantial. Urine, being an easily accessible and naturally excreted biofluid, eliminates the need for invasive procedures such as blood draws or biopsies, thereby reducing patient discomfort, risk of infection and cost of screening. When enhanced with AI-driven analysis, Neikuri has the potential to serve as a rapid, low-cost and accessible preliminary diagnostic tool—particularly in resource-limited settings where advanced medical infrastructure may be unavailable. By enabling early detection of imbalances or disease tendencies through pattern recognition, such a system could assist clinicians in triaging patients, monitoring chronic conditions or integrating complementary assessments alongside conventional medical evaluations. This fusion of traditional knowledge and modern technology paves the way for safer, culturally resonant and more inclusive diagnostic solutions in public health and truly contributing to the Indian Knowledge System.enImage ProcessingNeural NetworksMedicine DiagnosisIndiatext::thesis::doctoral thesis