This paper explores the application of deep learning algorithms in the accurate classification of tumours in medical images, addressing the critical need for precise cancer diagnosis. It reviews the utilization of various deep learning models, including Convolutional Neural Networks (CNNs), U-Net, Residual Networks, and others, in classifying tumours across different imaging modalities. The significance of deep learning in early detection, localization, and treatment planning is emphasized. The challenges and opportunities in this field are discussed, ranging from data availability and interpretability to data augmentation and interdisciplinary collaboration. By addressing these challenges, deep learning can significantly advance in cancer diagnostics, improving patient care.