Many clinical trials evaluating cancer treatments require patients to undergo testing for specific molecular markers as part of eligibility screening, typically using immunohistochemistry or sequencing. Because relatively few patients may carry a required marker, trial investigators often test large numbers of patients to identify the few who may ultimately qualify for enrollment.
Pathology laboratories routinely produce hematoxylin-and-eosin (H\&E) slides during cancer diagnosis. Pathology foundation models-large neural networks pretrained on millions of histology images-have shown promise in predicting molecular characteristics from these slides. Researchers can use these models to build classifiers that predict specific molecular markers and prioritize patients for confirmatory testing.
This study evaluates FATHOM (Facilitating Accrual through Tumor Histology and Omics Matching), an autonomous research system powered by large multimodal models. Its agents read registered clinical trial records, identify molecular markers used as enrollment criteria, build prediction models using pathology foundation models, select the individual models or model combinations that best meet prespecified criteria, set their decision thresholds, and determine whether to deploy them. Together, a prediction model, its decision threshold, and the decision to deploy it constitute an AI prediction policy.
Before FATHOM runs, the investigators preregister the clinical trial records that its agents may read, the cutoff date that defines which trial information they may use, the rules governing the agents, and the analysis plan. The system timestamps and locks each policy immediately after an agent produces it. The investigators then apply the policies to archived patient slides and compare their predictions with existing molecular marker results.
The primary outcome is the proportion of prespecified evaluation scenarios in which an agent-generated policy, compared with universal molecular testing, either enriches the population selected for confirmatory testing with marker-positive patients or safely spares patients from confirmatory testing while meeting prespecified performance criteria.
This study analyzes existing pathology images and clinical trial records only. It does not enroll or contact patients, influence patient care, or affect participation in any clinical trial.
Patients with a histologically confirmed cancer
Availability of relevant molecular profiling results
At least one diagnostic hematoxylin and eosin (H&E) whole-slide image
Poor-quality or unreadable slides, assessed independently of model output
Patients whose slides were used to train a policy's classifier, for that policy's evaluation
Patient records from archived, multi-institutional cohorts of patients with histologically confirmed cancers, relevant molecular profiling results, and at least one diagnostic hematoxylin and eosin (H\&E) whole-slide image. For the evaluation of a given policy, patients whose slides were used to train that policy's classifier are excluded. No patients are enrolled or contacted; enrollment counts refer to patient records analyzed.
Cohort
Retrospective
Not listed
Proportion of prespecified evaluation scenarios in which an AI-generated deployment policy demonstrates effective screening enrichment or rule-out performance
An evaluation scenario consists of one molecular marker evaluated in one study cohort. For each prespecified scenario, the study assesses whether the AI system generates a policy that either prioritizes patients more likely to carry the marker for confirmatory testing or identifies patients who may safely be spared testing, compared with testing everyone. The outcome is the proportion of scenarios in which the policy meets these performance criteria.
Per-scenario performance of each AI-generated deployment policy
For each evaluation scenario, the study reports sensitivity, negative predictive value, positive predictive value, the proportion of patients spared confirmatory testing, and the applicable enrichment or depletion ratio with its confidence interval.
Temporal generalizability for trials first posted on or after January 1, 2026
Among trials first posted on or after January 1, 2026 that require a molecular biomarker for enrollment, the study evaluates: (1) the proportion of biomarkers and trials for which the agentic AI screening approach is useful; and (2) the estimated number of patients who would benefit from AI-guided screening compared with universal molecular testing. Estimates are based on model performance and biomarker prevalence observed in the evaluation cohorts.
Temporal performance for trials first posted on or before December 31, 2025
Among trials first posted on or before December 31, 2025, that require a molecular biomarker for enrollment, the study evaluates: (1) the proportion of biomarkers and trials for which the agentic AI screening approach is useful; and (2) the estimated number of patients who would benefit from AI-guided screening compared with universal molecular testing. Estimates are based on model performance and biomarker prevalence observed in the evaluation cohorts.
Proportion of evaluation scenarios in which a non-default AI-generated deployment policy demonstrates effective screening enrichment or rule-out performance
Among scenarios in which the deployed policy is not the default strategy of testing everyone, the study assesses whether the AI system generates a policy that either prioritizes patients more likely to carry the marker for confirmatory testing or identifies patients who may safely be spared testing, compared with testing everyone. The outcome is the proportion of these scenarios in which the policy meets these performance criteria.
Click on the lead sponsor to view all of their trials.
This trial is not recruiting at the moment. You can still explore other options:
