Artificial Intelligence, Digital Pathology, and Data-Driven Oncology
Session Overview
Artificial intelligence is transforming oncology by unlocking patterns within complex, high-dimensional data—from radiological images and whole-slide histopathology to electronic health records and genomic sequences. This convergence of computational power and clinical data promises to augment diagnostic accuracy, predict therapeutic response, and optimize personalized treatment planning. This session brings together computational scientists, clinical oncologists, pathologists, and ethicists to critically evaluate the current state of AI tools, their pathway to clinical validation, and the practical and philosophical challenges of integrating them into cancer care.
Why This Session Matters Now
The proliferation of AI publications has far outpaced the adoption of AI tools in routine clinical practice, highlighting a critical implementation gap. While algorithms demonstrate impressive performance in silico, their real-world utility depends on rigorous clinical validation, seamless workflow integration, and clear demonstration of improved patient outcomes. This session addresses the essential transition from algorithmic development to responsible deployment, focusing on the multidisciplinary effort required to build transparent, equitable, and clinically actionable AI systems that earn the trust of oncologists and patients.
Key Scientific and Clinical Themes
AI in Cancer Diagnosis, Imaging, and Radiomics
Examination of deep learning applications in medical imaging (CT, MRI, PET) for tumor detection, characterization, and segmentation, and the extraction of quantitative radiomic features for prognostic and predictive modeling.
Digital Pathology and Computational Histopathology
Analysis of AI-powered tools for whole-slide image analysis, including tumor grading, identification of specific morphologic patterns, quantification of tumor-infiltrating lymphocytes, and prediction of molecular alterations from H&E stains.
Predictive Modeling for Treatment Response and Outcomes
Focus on the development and validation of multivariable models that integrate clinical, genomic, and imaging data to forecast individual patient prognosis, likelihood of response to specific therapies, and risk of recurrence or toxicity.
Clinical Decision Support Systems in Oncology
Discussion of how AI can be operationalized at the point of care, from recommending guideline-concordant therapy based on real-world evidence to prioritizing clinical trial matches and managing supportive care.
Real-World Data, Big Data, and Oncology Informatics
Exploration of the infrastructure and methodologies required to generate learning healthcare systems, including data harmonization, federated learning approaches, and the use of real-world evidence to complement clinical trial data.
Ethical, Regulatory, and Validation Challenges in AI Oncology
Critical appraisal of the barriers to clinical adoption, including algorithm bias and fairness, explainability (“black box” problem), regulatory pathways (FDA, CE marking), data privacy, and the necessity for prospective clinical utility studies.
Nature of Research in This Field
Research in oncology AI is characterized by a dichotomy: a vast body of early-stage, single-institution proof-of-concept studies (reflected in narrative reviews) and a smaller but growing pipeline of rigorous multi-center validation trials. The field requires unprecedented collaboration between computer scientists and clinical domain experts. A significant portion of the literature is dedicated to methodological standards, benchmarking, and addressing the sociotechnical challenges of implementation, reflecting a maturing focus on translational impact over purely technical innovation.
Who Should Attend
This session is designed for:
- Computational oncologists, radiomics scientists, and digital pathologists
- Medical oncologists, radiologists, and surgical oncologists interested in AI tools
- Data scientists, machine learning engineers, and biomedical informaticians
- Regulatory science professionals and health technology assessment experts
- Clinician-leaders involved in implementing digital health solutions
Session Perspective
The ultimate test of an oncology AI model is not its AUC on a retrospective dataset, but its ability to reliably improve a clinical decision for a specific patient at the bedside. This session provides a platform to move beyond the hype cycle and focus on the engineering of trust. By connecting technical advances in model development with the rigorous demands of clinical evidence generation and ethical practice, the discussion aims to chart a responsible and practical path for data-driven intelligence to become a foundational, trusted component of modern oncology.
If your research aligns with this session, we invite you to submit an abstract for consideration.
