AI may help diagnose pancreatic cancer earlier from CT scans


A study published in Gastroenterology finds that radiomics-based machine learning models can detect pancreatic cancer on prediagnostic CT scans much earlier than current clinical diagnostic methods.

“Pancreatic cancer is a deadly disease and one of the leading causes of cancer-related death,” says Ajit Goenka, MD, diagnostic radiologist at Mayo Clinic and lead author of the study.

Dr. Goenka says that while early detection improves the chances of successful treatment, standard imaging cannot detect pancreatic cancer at an early stage.

“Up to 40% of small pancreatic cancers may not show up on standard imaging. Therefore, the majority of patients present with advanced, non-curable disease,” says Dr. Goenka.

For this reason, Dr. Goenka and his colleagues sought to integrate artificial intelligence (AI) into radiological screening to detect pancreatic cancer at an earlier and more curable stage. “We found that AI models can detect cancer from a normal-appearing pancreas on CT scans months before cancer symptoms, even when the disease was beyond the scope of radiologists’ perception.”

For the study, the researchers computed the imaging signature of early cancer from the scans. Prediagnostic CT scans are CT scans that were performed for unrelated indications between three months and three years before the onset of the cancer.

Then they used a group of age-matched control subjects who did not develop pancreatic cancer during the three years of follow-up. Expert radiologists then segmented the pancreas on the CT scans of both groups and extracted and computer-quantified measures of pancreatic tissue heterogeneity.

Next, the researchers built advanced machine learning models that could predict future pancreatic cancer risk at a median time of 386 days, a range of 97 to 1,092 days, before clinical diagnosis with accuracies ranging from 94 % to 98%.

“In comparison, radiologists were unable to reliably differentiate patients who developed cancer from those who had a normal pancreas,” says Sovan Mukherjee, Ph.D., senior data science analyst at Dr. Goenka’s team and first author of the study. . “We also tested our AI models against variations in image noise, scanner patterns, image acquisition protocols, and post-processing settings, and found that they did not were not affected by these variations.”

Dr. Goenka says this level of testing is necessary for the potential deployment of this technology in clinical practice. Finally, the researchers validated the high specificity – 96.2% – of the AI ​​model on an open-source CT dataset to further increase the reliability of the AI ​​methodology.

“Our study demonstrates that artificial intelligence can identify asymptomatic people who may harbor occult cancer at a stage where surgical cure is possible,” says Dr. Goenka. “These findings may help overcome one of the major barriers to improving survival for patients with pancreatic cancer.”

Dr Goenka says a large, prospective clinical trial – the Early Detection Initiative (NCT04662879) sponsored by the Pancreatic Cancer Action Network – is underway to assess the impact of a pancreatic cancer screening strategy using CTs. among 12,500 participants. The trial is led by Suresh Chari, MD, gastroenterologist emeritus at the Mayo Clinic. Dr. Goenka’s team is exploring the option of prospective validation of their AI models on CTs to be performed as part of the EDI trial.

Reference: Mukherjee S, Patra A, Khasawneh H, et al. Radiomics-based machine learning models can detect pancreatic cancer on prediagnostic CT scans within a substantial time before clinical diagnosis. Gastroenterology. July 1, 2022. doi:10.1053/j.gastro.2022.06.066

This article was republished from the following documents. Note: Material may have been edited for length and content. For more information, please contact the quoted source.

Source link


Comments are closed.