Jul 2026· AI in Precision Oncology· Vol 3, pp. 107 - 110· 0 citations· 10 references
TL;DR
This commentary describes how FL is evolving into an essential component across all stages of the precision oncology process, including areas such as radiomics, digital pathology, genomics, multi-omics analysis, molecular tumor board decision-making, patient matching in clinical trials, and development of multimodal AI algorithms.
Abstract
Precision oncology, an evolving branch of medical science, is increasingly dependent on the application of AI techniques to aid accurate diagnosis, molecular profiling, proper therapy selection, and outcome prediction in cancer patients. However, several problems such as fragmented health care data, institutional barriers, and privacy restrictions related to the use of patient data continue to impede the development of clinically applicable and generalizable AI models. Federated learning (FL) is a newly developed paradigm that enables collaborative creation of AI models without compromising data privacy. In this commentary, we describe how FL is evolving into an essential component across all stages of the precision oncology process, including areas such as radiomics, digital pathology, genomics, multi-omics analysis, molecular tumor board decision-making, patient matching in clinical trials, and development of multimodal AI algorithms. We also address the issue of transitioning from a model with strong technical performance to one that demonstrates good clinical performance by considering aspects of prospective validation, integration into workflows, model interpretability, cybersecurity, governance, and regulation. FL should be thought of as crucial infrastructure rather than just an efficient machine-learning technique for privacy-sensitive environments.
Precision oncology has evolved from single-gene biomarker testing toward multimodal molecular and clinical profiling; however, most therapeutic decisions remain based on static baseline assessments that inadequately capture tumor evolution, treatment response, and emerging resistance. In this perspective, we propose a forward-looking framework that integrates federated learning, pharmacogenomic digital twins, and hybrid quantum-classical optimization to support the development of adaptive, privacy-preserving precision oncology systems. The framework enables collaborative model training across institutions without sharing raw patient data, thereby addressing major barriers associated with data fragmentation and privacy regulations. Patient-specific digital twins serve as continuously evolving computational representations that integrate longitudinal multiomics, imaging, pathology, and clinical information to simulate disease trajectories, estimate therapeutic response, and anticipate resistance patterns. Federated learning allows these models to benefit from geographically distributed patient cohorts while maintaining data sovereignty and institutional privacy. In contrast to near-term deployable components such as federated learning and digital twin modeling, quantum computing is presented as a future-oriented computational strategy that may assist selected combinatorial optimization tasks, including treatment selection, dose optimization, and scheduling, through hybrid quantum-classical workflows. Rather than assuming immediate clinical utility or quantum advantage, the framework emphasizes realistic translational pathways that acknowledge current limitations in quantum hardware, scalability, validation, and regulatory readiness. We further discuss technical feasibility, challenges associated with heterogeneous clinical data, privacy considerations, validation requirements, and regulatory pathways for adaptive AI-enabled clinical decision-support systems. By providing a formal conceptual architecture and translational roadmap, this perspective outlines how federated artificial intelligence, continuously learning digital twins, and future quantum-assisted optimization may collectively contribute to the next generation of adaptive precision oncology.
Abstract The integration of artificial intelligence (AI) in digital pathology has shown significant promise in advancing cancer diagnostics, grading, and treatment response prediction. However, widespread development and deployment of robust AI models face critical challenges due to data silos, privacy concerns, and the need for large-scale multi-institutional datasets. Federated Learning (FL) presents a transformative approach by enabling collaborative model training across hospitals without direct data sharing. In this review, we summarize recent developments in FL as applied to digital pathology, highlight pioneering use cases, and explore the technical, regulatory, and ethical hurdles. We discuss how FL can enable scalable, privacy-preserving AI models, and outline future directions for standardizing and validating FL-based approaches in clinical workflows.
Abstract Artificial intelligence (AI) is increasingly embedded across oncology workflows, with applications spanning initial assessment, diagnostic work-up, treatment selection, longitudinal monitoring, and survivorship care. A clinically oriented synthesis of these use cases is needed to guide real-world adoption and highlight implementation challenges. A narrative review of recent clinical, translational, and health-services literature was conducted, focusing on AI tools deployed in routine or near-term oncology practice. The manuscript organizes evidence along the cancer care continuum, from first presentation to outcome prediction, and integrates ethical, legal, and regulatory perspectives. AI systems now support ambient documentation and natural language processing (NLP) for initial consultations, advanced imaging and digital pathology for diagnosis, and multimodal decision support for systemic therapy, radiation, and surgery. Additional applications include toxicity and adverse-event prediction, real-time symptom and liquid biopsy monitoring, risk-adapted follow-up, and survivorship risk stratification, although prospective validation and interoperability remain uneven. AI has moved from proof-of-concept to a practical adjunct to oncology decision-making, with demonstrable potential to enhance precision, efficiency, and patient experience across the cancer continuum. Realizing this promise will require validated, explainable, and equitable systems, robust data governance, and deliberate design of human–AI collaboration within everyday oncology practice. AI would not be a replacement for human expertise, but a pivotal tool to aid cancer patient care.
Ganesh H. Divekar, Bharat Bhosale· Indian Journal of Medical an...· 0 citations
Overall, AI is becoming an integral component of modern oncology, particularly radiation oncology, and its successful integration into routine clinical practice will require robust validation, transparent governance, equitable implementation, and continued clinician oversight to ensure safe, effective, and patient-centred cancer care.
K. Rastogi· The Rise of Artificial Intel...· 0 citations
Rare malignancies, defined by an incidence of <6 cases per 100,000 individuals per year, collectively constitute 25% of the global cancer burden. However, progress is hindered by the "diagnostic odyssey" and structural barriers that concentrate clinical trials in urban academic centers, excluding 32 million rural Americans and leading to premature termination of up to 40% of rare cancer trials. This study evaluates the integration of federated learning architecture and decentralized master protocols to overcome these systemic challenges. We present technical benchmarks from the Cancer AI Alliance (CAIA), which utilizes the NVIDIA FLARE™ ecosystem to train predictive models on de-identified data from over one million patients across multiple institutions. This decentralized approach circumvents privacy constraints, allowing for the identification of rare genomic signatures and treatment resistance patterns with a tenfold increase in speed. Furthermore, we analyze data from the TCF-001 TRACK trial, a fully remote precision genomics study. Results demonstrate that utilizing multidisciplinary Virtual Molecular Tumor Boards (VMTBs) to interpret comprehensive genomic profiling (CGP) achieves high therapeutic matching scores (≥ 50%) for 41% of participants, significantly correlating with improved progression-free and overall survival in refractory cohorts. Disease-specific analysis highlights the impact of targeted PKC inhibition with darovasertib in primary uveal melanoma, achieving a 95% eye preservation rate among responders and establishing a vision-sparing neoadjuvant paradigm. In chordoma, the structural elucidation of the driver Brachyury (TBXT) in early 2025 has transitioned the field from "undruggable" hypotheses to fragment-based ligand optimization and the opening of first-in-human TCR T-cell trials in 2026. Complementary advancements in ultrasensitive liquid biopsy (superRCA) allow for the detection of minimal residual disease at tumor fractions below 0.01%, facilitating early interceptive treatment of relapse. Together, these technological and methodological shifts signal a transition toward "Precision Medicine 2.0", where spatial multi-omics and AI-driven stratification ensure diagnostic equity and therapeutic innovation for all patients, regardless of cancer rarity. Methods: This research synthesized evidence from the TCF-001 TRACK remote clinical trial, longitudinal data from the OptimUM-09 uveal melanoma program, and the technical implementation benchmarks of the CAIA federated learning platform. Generative AI was used in the development of this abstract to synthesize complex multi-institutional data and ensure stylistic alignment with 2026 AACR submission guidelines.
Debanjan Gangopadhyay, Debanjan Gangopadhyay. Radical collaboration in rare oncology: Accelerating therapeutic validation through federated learning and remote precision master protocols [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Breaking Barriers in the Fight against Rare Cancers; 2026 Jul 18-20; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2026;86(14_Suppl):Abstract nr B034.
Debanjan Gangopadhyay· Cancer Research· 0 citations
Breast cancer (BC) ranks among the most common malignant tumors affecting women globally. The essence of precision medicine (PM) lies in “delivering the right treatment to the right patient at the right time.” With advancements in artificial intelligence (AI) technologies such as deep learning (DL), breakthroughs have been achieved in analyzing data ranging from imaging to multi-omics. We review the latest applications and challenges of AI in PM for BC, offering insights for clinical practice and research. We also present an AI integration framework covering the entire BC care continuum. The framework systematically integrates multiple components, including imaging diagnosis, digital pathology, multi-omics analysis, treatment response prediction, surgical decision-making, clinical decision support, and clinical translation, thereby revealing the hierarchical mechanisms through which AI contributes to the precision management of BC. This paper reviews how AI can enable precise management of BC patients across different temporal and biological scales by collecting different types of data. Specifically, this encompasses precision prevention, diagnosis, and clinical management. It also highlights current research gaps and challenges, such as algorithmic bias, dataset comprehensiveness, and model interpretability. Ultimately, the paper offers valuable insights into the integration of AI throughout the entire process of precision medical management for BC patients.
