Sycai Medical at the XXIV SEDIA Congress: validating AI for the detection of Pancreatic Cystic Lesions
Last October, our collaboration with Vall d’Hebron University Hospital reached a new milestone: at the XXIV SEDIA Congress, held in Bilbao, Dr Andreu Antolín presented the results of a clinical study evaluating the performance of our deep learning model when assisting radiologists in detecting pancreatic cystic lesions (PCLs) on CT scans.
PCLs are often subtle and incidental, and therefore easy to overlook. However, their correct identification and long-term follow-up can transform clinical decision-making. This makes the pancreas a particularly challenging organ in which AI could have a significant impact.
What did the study look at?
The research team analyzed over a hundred CT scans of patients with confirmed cystic lesions and controls using a deep neural network trained for segmentation and detection. The focus was simple: could SYCAI MEDICAL reliably detect and localize these PCLs, even when they were small or difficult to see?
Key highlights:
- Strong overall performance: SYCAI MEDICAL accurately detected most pancreatic cysts (91.4%), achieving a sensitivity of 93% and an F1 score of 96%.
- Solid performance on small lesions: sensitivity remained high (up to 83% and F1 score of 91%) even for lesions under 10 mm — a notoriously challenging category for radiologists.
- Reliable anatomical localization: the model correctly localized 72% of cases within the pancreas.
These results do not replace expertise; they enhance it. In a field where early detection can truly transform outcomes, these improvements are by no means trivial.
This validation study from Vall d’Hebron demonstrates the valuable support that AI can provide in the detection and monitoring of pancreatic cystic lesions in CT imaging. With its robust performance across lesion sizes and anatomies, the study paves the way for the integration of reliable, patient-centred AI into routine radiological practice.
For Sycai Medical, this is another promising step forward, reinforcing our mission to bring clinically impactful AI into routine practice.
