Cedars-Sinai researchers are using an advanced AI tool to aid in the early detection of the most commonly fatal form of pancreatic cancer, pancreatic ductal adenocarcinoma, which holds a five-year survival rate under 10%. Although it has been proven that detection in initial stages can lift that rate by up to 50%, there is a distressing lack of reliable methodology for early diagnosis. The Cedars-Sinai team’s AI innovation, according to a study published in the Cancer Biomarkers journal, uses imaging data collated from at most three years prior to official diagnosis to detect pancreatic ductal adenocarcinoma.
“This AI tool was able to capture and quantify very subtle, early signs of pancreatic ductal adenocarcinoma in CT scans years before occurrence of the disease,” said Dr. Debiao Li, the study's senior author and Director of Cedars-Sinai’s Biomedical Imaging Research Institute. “These are signs that the human eye would never be able to discern.” The tool was developed to evaluate CT images for textural changes on the surface of the pancreas, which are indicative of the underlying molecular alterations associated with the growth of pancreatic cancer.
Training the tool for this purpose entailed feeding the AI “normal” pancreatic CT scan sets that had been sampled between six months to three years before fewer than 40 patients were diagnosed with the cancer. Comparisons between cancerous and benign scans, the latter taken from the same number of patients, helped the AI learn to discern between the two. The AI model ended up reaching 86% accuracy in predicting eventual tumor development in individuals.
