AI for Early Pancreatic Lesion Detection: Highlights from the Congrés de Radiòlegs de Catalunya
Last month, we had the pleasure of participating in the X Congrés Nacional de Radiòlegs de Catalunya, held in Barcelona. On the opening day, Dr Montserrat Duh presented the results of a clinical study conducted by the Radiology Department at Hospital de Mataró (Consorci Sanitari del Maresme) under the direction of Dr Mayte Fernández Planas, in collaboration with the Sycai Medical team. The presentation focused on one of the biggest challenges in abdominal radiology today: the early detection of pancreatic findings — both benign and malignant — using artificial intelligence.
Pancreatic cancer is one of the deadliest cancers and is often detected too late. Cystic and solid pancreatic findings may be silent or subtle in the early stages and manually reviewing multiple CT scans for follow-up can be time-consuming and error-prone.
Our goal? To support radiologists by providing them with a fully automated algorithm that can detect, segment and monitor these findings across multiple time points, saving time, improving accuracy and helping to spot changes that might otherwise go unnoticed.
The study included 103 patients with pancreatic findings and a control group without. The findings were categorized as either cystic (n = 38) or solid (n = 29), including adenocarcinoma and neuroendocrine tumors.
Key results:
✔️Overall sensitivity: 91%
✔️Specificity: 88%
✔️Almost perfect agreement with expert radiologists (κ = 0.98)
✔️Our algorithm retrospectively detected a finding in the initial CT scan of a patient with early adenocarcinoma, which had not been detected by the radiologist at the time.
These results are particularly significant for slowly evolving, subtle lesions — cases where early detection can be life-changing.
Our results support the value of AI as a clinical decision support tool, particularly regarding longitudinal assessment and follow-up. We would like to thank the Hospital de Mataró team for their trust and collaboration, continuing to push forward with the integration of AI into radiologist’s workflow, enhancing their work, speeding up processes and improving patient outcomes.
