Computer-aided diagnosis (CAD) tools can help you better and faster figure out the risk level of tumours from radiology images. Using these tools to describe tumours can also help with non-invasive cancer staging, prognosis, and personalised treatment planning as part of precision medicine. This paper presents both supervised and unsupervised machine learning methodologies aimed at enhancing tumour characterisation. Our initial methodology relies on supervised learning, where we illustrate substantial improvements achieved through deep learning algorithms, specifically by employing a 3D convolutional neural network and transfer learning. Inspired by the interpretations of scans by radiologists, we demonstrate the integration of task-dependent feature representations into a CAD system through a graph-regularized sparse multi-task learning framework. In the second approach, we examine an unsupervised learning algorithm to mitigate the scarcity of labelled training data, a prevalent issue in medical imaging applications. Based on what we learned from label proportion methods in computer vision, we suggest using a proportion-support vector machine to describe tumours. We also want to know if
Dr.M.Sukesh et al. (Wed,) studied this question.