Deep learning in oncology: Transforming cancer diagnosis, prognosis, and treatment

• Published in: Emerging trends in drugs, addictions, and health Vol. 5; p. 100171
• Main Author: Reis, Tiago Cunha
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.12.2025; Elsevier
• Subjects: Artificial Intelligence; Clinical Trials; Deep Learning; Drug Design; Imaging Reports; Multiomics; Oncology; Precision Medicine; Psychiatric/Mental Health; Artificial Intelligence; Drug Design; Imaging Reports; Oncology; Multiomics; Clinical Trials; Deep Learning; Precision Medicine
• ISSN: 2667-1182; 2667-1182
• DOI: 10.1016/j.etdah.2025.100171

Deep learning (DL) has emerged as a transformative force in oncology, offering unprecedented capabilities in cancer diagnosis, treatment planning, and prognosis. The integration of DL models with vast and complex datasets, including genomic, transcriptomic, and imaging data, has paved the way for more precise and personalized cancer care. In particular, DL's application in drug efficacy and toxicity prediction is gaining traction, addressing the critical challenge of high drug failure rates in clinical development. By leveraging large datasets and sophisticated algorithms, DL models can predict drug responses and optimize treatment strategies, ultimately improving patient outcomes. Additionally, DL-driven automation in medical imaging processing and report generation is revolutionizing radiology, enhancing diagnostic accuracy and consistency. This review explores the current advancements in DL applications across various aspects of oncology, emphasizing the potential of AI-driven tools to enhance the accuracy, efficiency, and personalization of cancer care. Despite the significant progress, challenges such as model validation, ethical considerations, and the need for transparent AI systems remain. Addressing these challenges will be crucial in realizing the full potential of DL in transforming oncology practices. The graphical abstract underscores AI's transformative role in oncology, focusing on three key areas: multiomics integration for personalized cancer treatment, predictive modeling for drug efficacy and toxicity, and enhanced radiologic image interpretation. AI-driven tools facilitate the tailoring of treatments to individual patients by analyzing complex multiomics data, while also optimizing therapeutic strategies through accurate predictions of drug efficacy and potential toxicity. Additionally, AI significantly improves the speed and detail of radiologic interpretations, supporting more precise diagnoses and treatment planning in cancer care. [Display omitted]