Dynamic AI Tissue Analysis For Enhanced Medical Diagnosis using TSDM

The aim of the study is to propose TSDM-Net, a model that enhances tissue analysis accuracy through adaptation to various tissue types, and compare its performance with conventional CNN models. In this research, there are two groups. Group 1 is the CNN with attention based on the CAMELYON16 dataset, with annotated lymph node WSIs, while Group 2 is the TSDM-Net with the Break-His dataset, with benign and malignant samples of breast cancer. Preprocessing included color normalization, resizing to 512×512 pixels, and augmentation processes such as rotation, flipping, and contrast changes. The TSDM AI model performed better than the CNN model in separating healthy and diseased tissues, with a 94% accuracy and substantial improvement in precision, recall, and F1-score over 20 epochs of training. The precision (0.95) of the TSDM improved from 0.8000 to 0.9000, recall (0.94) from 0.7800 to 0.8800, and F1-score (0.945) from 0.7900 to 0.8900 at a significance level of 0.0056. In this work, it is noted that the combination of AI tissue analysis and Time-Series Data Mining (TSDM) greatly improves medical diagnostics, especially in the detection of early diseases, by enhancing the analysis and interpretation of intricate tissue data.