Developing individual, disease-specific AI models for less common and rare diseases remains a challenge due to data scarcity and a long tail of atypical diagnoses for many indications. An alternative approach, known as anomaly detection, uses common disease data to train AI models on what “normal” looks like, so that anomalies can be flagged.

Aignostics trained its foundation model, RudolfV, using a proprietary approach that enhances generalizability and sought to test how RudolfV performed on an anomaly detection task. When benchmarked against other foundation models, RudolfV was better able to recognize rare disease anomalies, highlighting its potential for improving diagnostic accuracy in rare cancers.

Please fill out the form below to receive the full case study. Please note we are unable to respond to requests from personal email addresses.