Blood Cancer Prediction System using Tissue Image Dataset

The histopathological examination of tissue biopsies is the gold standard for diagnosing blood cancers like leukaemia and lymphoma. This process, however, is highly dependent on the expertise of pathologists, can be time-consuming, and is susceptible to inter-observer variability. This paper presents a deep learning-based framework for an automated blood cancer prediction system using a tissue image dataset. We propose a system that leverages a Convolutional Neural Network (CNN) to analyse digital images of bone marrow or lymph node biopsies and classify them as either benign or malignant. The methodology employs transfer learning with a pre-trained CNN architecture, fine-tuned on a curated dataset of annotated histopathology patches. This approach is designed to learn the intricate morphological features and cellular patterns indicative of haematological malignancies. The system aims to serve as a robust and efficient decision-support tool, assisting pathologists by providing rapid, objective, and accurate preliminary classifications, thereby improving diagnostic workflow and consistency.