This paper provides a comprehensive overview of diverse techniques employed in cancer research using various datasets and modalities. A multi modal fusion paradigm is introduced for tumor response and resistance prediction, utilizing the TCGA dataset and achieving notable performance. Transfer learning is employed to analyze cancer diagnosis and treatment, attaining significant accuracy improvements for multiple architectures. Radiomic analysis is used to cluster patients based on CT scan images, demonstrating substantial clustering accuracy. Additionally, a deep learning pipeline technique is proposed for more accurate registration results in prostate cancer detection. Radiomic features and feed-forward networks are applied for histological cancer prediction in central lung, achieving high AUC values. A Multi-modal Multi Scale Attention Model (MMAM) can predict lung cancer stages using TCGA data. The review also discusses the AJIVE technique for exploratory analysis and an InceptionNet-v3 tool for classifying breast cancer genomic biomarkers from histopathological images, achieving promising results in diagnosing cancer. This compilation offers valuable insights into cutting-edge techniques for cancer research across various modalities and datasets.