Recently, an article published on Nature Medicine gave us insight into the possibilities of using AI for cancer detection.
The article discusses a deep learning approach, named PANDA, for detecting and classifying pancreatic lesions.
The article highlights that Pancreatic ductal adenocarcinoma (PDAC), a deadly solid malignancy, is often detected late and at an inoperable stage. It mentions that Non-contrast computed tomography (CT), which is routinely performed for clinical indications, offers potential for large-scale screening.
The researchers developed PANDA to detect and classify pancreatic lesions with high accuracy via non-contrast CT. PANDA was trained on a dataset of 3,208 patients from a single center. It achieved an area under the receiver operating characteristic curve (AUC) of 0.986–0.996 for lesion detection in a multicenter validation involving 6,239 patients across 10 centers.
PANDA outperformed the mean radiologist performance by 34.1% in sensitivity and 6.3% in specificity for PDAC identification. In a real-world multi-scenario validation consisting of 20,530 consecutive patients, PANDA achieved a sensitivity of 92.9% and specificity of 99.9% for lesion detection.
The article concludes by stating that PANDA, when used with non-contrast CT, showed non-inferiority to radiology reports (using contrast-enhanced CT) in differentiating common pancreatic lesion subtypes. This suggests that PANDA shows promise in the early detection of pancreatic cancer using non-contrast CT scans.