Dana-Farber AI traces unknown cancers back to their source

Researchers from Dana-Farber Cancer Institute have created an artificial intelligence tool designed to trace the origin of cancer in challenging cases. Even with standard diagnostic approaches involving imaging scans, biopsies, and pathology reports, pinpointing the source of malignancy remains elusive in approximately 3% to 5% of patients.

Referred to as cancers of unknown primary origin, these instances involve metastatic cancer that has spread from a hidden location. Typically, treatment options are limited for these patients, as therapies are tailored to specific tumour types. The AI tool, called OncoNPC, utilises tumour DNA sequencing data to predict the cancer's source. The model, based on data from over 36,400 patients, uses machine learning to classify patients with difficult-to-diagnose cancers. According to the study published in Nature Medicine, OncoNPC accurately predicted the origin of around 80% of tumours, potentially expanding the pool eligible for precision treatments.

The researchers found that the model's high-confidence predictions were correct about 95% of the time. This AI tool could provide valuable insights for treatment decisions, including precision medicine. Future plans involve enhancing the model's performance by incorporating additional data, such as pathology reports, and exploring its integration with other diagnostic methods in the field of cancer treatment.

Click here to read the original news story.