Mayo Clinic AI Detects Pancreatic Cancer Years Before Diagnosis

Mayo Clinic researchers developed an artificial intelligence model that detects pancreatic cancer signals up to three years before a clinical diagnosis [1, 2].

This advancement addresses a critical gap in oncology, as more than 85% of patients are currently diagnosed only in advanced stages ^[1]. Early detection is vital for a cancer known for its high mortality rate and limited treatment windows.

Based in Rochester, Minnesota, the research team included Dr. Ajit Goenka [1, 3]. The AI model analyzed nearly 2,000 routine CT scans to identify subtle patterns invisible to the human eye ^[1]. These scans are often performed for other medical reasons, meaning the AI can find indicators of malignancy in patients who do not yet show symptoms.

Data regarding the lead time of these detections vary across reports. Some findings indicate the AI can identify the disease up to three years in advance [1, 2], while other data suggests a detection window of 475 days before a formal diagnosis ^[3]. The model demonstrated a sensitivity of 73% in identifying the disease ^[3].

Pancreatic cancer remains one of the most difficult malignancies to treat because it is often asymptomatic until it has spread. By utilizing routine imaging, which is already common in healthcare settings, the model aims to shift the diagnosis from late-stage palliative care to early-stage intervention.

The researchers used the large dataset of scans to train the algorithm to recognize the earliest morphological changes in the pancreas. This process allows the AI to flag high-risk patients for more frequent monitoring or immediate biopsy, potentially increasing the survival rate for those who would otherwise be diagnosed too late [1, 2].