Clinical bottom line
Artificial intelligence (AI) applied to liquid biopsy data represents a promising frontier in pathology for predicting key pathological features across various cancers. Current evidence demonstrates AI’s potential to enhance sensitivity and specificity in detecting tumor-derived biomarkers such as circulating tumor DNA (ctDNA), microRNAs, and exosomes, facilitating non-invasive tumor characterization and monitoring. However, most studies remain exploratory or early clinical validations, with heterogeneity in AI models and biomarker panels. Integration of AI-driven liquid biopsy analyses into routine pathology practice could improve diagnostic accuracy and personalized treatment decisions but requires further prospective validation and standardization (Mishra et al., 2024; Boukovala et al., 2024; Chenchula et al., 2026).
What the evidence shows
Mishra et al. (2024) review liquid biopsy applications in gallbladder carcinoma, highlighting AI algorithms that analyze ctDNA and protein signatures to predict tumor stage, histological subtype, and treatment response. They report improved diagnostic performance compared to conventional methods, though emphasize the need for larger cohorts.
Boukovala et al. (2024) provide a broad overview of liquid biopsy in clinical oncology, underscoring AI’s role in integrating multi-omic data from blood samples to identify pathological features such as mutation burden and tumor heterogeneity. They discuss machine learning models that classify cancer subtypes and predict prognosis with promising accuracy.
Chenchula et al. (2026) summarize current clinical evidence and future directions, noting that AI-enhanced liquid biopsy enables dynamic monitoring of tumor evolution and resistance mechanisms. They highlight challenges in data standardization and the necessity of robust validation frameworks.
Shigeyasu et al. (2023) focus on microRNAs as liquid biopsy biomarkers in gastrointestinal cancers, demonstrating that AI models can effectively discriminate malignant from benign conditions and correlate microRNA patterns with pathological aggressiveness.
OKTAY et al. (2026) discuss thyroid cancer, where AI applied to liquid biopsy data improves detection of genetic alterations and pathological risk stratification, potentially reducing the need for invasive tissue biopsies.
Caveats and uncertainty
Most AI models are trained on limited, retrospective datasets, risking overfitting and limited generalizability.
Variability in liquid biopsy assay platforms and biomarker selection complicates cross-study comparisons.
Standardized protocols for sample collection, processing, and data annotation are lacking.
Regulatory approval pathways for AI-based diagnostic tools remain evolving.
Clinical utility in diverse populations and cancer types requires further prospective validation.
Interpretability of AI algorithms and integration with existing pathology workflows pose practical challenges.
How this may change practice
AI-driven analysis of liquid biopsy data may complement traditional pathology by providing rapid, non-invasive insights into tumor biology.
Pathologists may increasingly rely on AI tools to interpret complex molecular data, enhancing diagnostic precision and personalized therapy selection.
Routine monitoring of tumor dynamics via liquid biopsy could enable earlier detection of treatment resistance and relapse.
Multidisciplinary collaboration among pathologists, bioinformaticians, and clinicians will be essential to implement AI-enabled liquid biopsy effectively.
Ongoing research and clinical trials should focus on validating AI models in prospective cohorts and establishing standardized workflows.
Regulatory and ethical frameworks must evolve to ensure safe, equitable deployment of AI diagnostics.