Axillary Lymph Node Metastasis prediction Using Deep Reinforcement Learning on Primary Tumor Biopsy Slides

The “Sentinel Lymph Node Biopsy” (SLNB) is a common practice for detecting metastasis in Breast Cancer patients. It has improved the accuracy of predicting “Axillary Lymph Node” (ALN) status, allowing some women to avoid “Axillary Lymph Node Dissection” (ALND). However, SLNB is an invasive technique that has the potential to induce problems such as “lymphedema” and numbness in the upper limbs. Furthermore, the frozen segment of the SLNB takes a significant amount of time, which extends the total duration of the surgery. In this research work, a Deep Reinforcement Learning (DRL) approach has been used to pre-operatively predict ALN involvement in early-stage Breast Cancer patients using Whole Slide Images (WSIs). The goal was to achieve accurate ALN involvement detection, which is crucial in determining the best treatment options and avoiding unnecessary surgery and related complications. A Convolutional Neural Network model based on Deep Learning was also used for ALN metastasis prediction. The results of the experimentation demonstrate that the DRL algorithm outperformed the CNN model in terms of accuracy, precision, recall, and F1-score, with values of 67.7121%, 65.1499%, 67.7121%, and 63.519% respectively.