Cancer Blood Disorder Detection in Blood Smear Images using ResNet50 based Deep Learning Model

This study provides a novel deep-learning framework for the diagnosis of cancer blood disorder from blood smear images using the ResNet50 architecture with customized modifications. The main contribution is to improve the pre-trained ResNet50 model's robustness and diagnosis accuracy in real-world medical situations by fine-tuning the customized layers. A balanced dataset of 2,465 blood images, evenly split between patients with healthy and unhealthy instances, was used to develop the model. Image preprocessing and image classification are two of the study's phases. Early detection of cancer blood disorder is achieved through classification using the combined 2D Iterative Convolutional Neural network algorithm and ResNet50 classifier. Advanced pre-processing techniques, segmentation techniques, texture analysis, and the capability of ICNN have all been combined to build a system that can achieve the best accuracy, precision, recall, sensitivity, and F1 Score of this model, measures were 97%, 97.49%, 97.5%, 97.25, and 97.35% respectively. As a performance test, to evaluate our system against current classifiers. In comparison to existing approaches, we created a comprehensive system for classifying and forecasting blood anomalies relevant to cancer. Our suggested approach's power and superiority were confirmed by the comparison study with other deep learning classifiers. This research work might significantly advance medical diagnostics by giving medical practitioners a formidable tool for the early and precise prediction of cancer blood disorder.