Pattern Recognition and Deep Learning Models for Cancer Detection

Abstract

Cancer, a leading cause of death worldwide, poses challenges due to late-stage detection and inaccurate imaging techniques. Precise and effective screening is crucial for early detection and treatment. Various low-cost and accurate imaging techniques are used, but difficulties arise when experts struggle to interpret certain image areas, leading to missed cancer diagnoses. To address this, computer-based software utilizing deep learning models and algorithms has been developed. Traditional approaches have evolved into computerized tools that analyze, diagnose, and predict symptoms. This study conducted a comparative analysis of three detection models of different frameworks for binary and multi-class image classification using image processing techniques. Despite their distinct inputs, model network architecture Convolutional Neural Networks (CNN). The data were split into sets: training, validation, and testing, and the evaluation involved learning curves of training and validation loss and accuracy as well as a comparison of training, validation, and test accuracies to check the model performances. The results showed the accuracy of all models of predicting unknown images of cancerous or non-cancerous. The Difference in input data and model learning rate could be the reason for variation in test results. Computer language, python, and the platform, Pycharm IDE have been used for tasks performed.