In a study (PANORAMA) reported in The Lancet Oncology, Alves et al found that an artificial intelligence (AI) program developed by the investigators was significantly better at detecting pancreatic ductal adenocarcinoma when applied to standard-of-care computed tomography (CT) scans than were experienced radiologists.
Study Details
The AI system was developed using CT scans in a cohort of 2,310 patients from four tertiary care centers in the Netherlands and the United States for training (n = 2,224) and tuning (n = 86) and in a sequestered cohort of 1,130 patients from five tertiary care centers in the Netherlands, Sweden, and Norway for testing. A subset of 391 patients from the testing cohort had CT scans read by 68 radiologists from 40 sites in 12 countries with a median of 9.0 years of experience (interquartile range = 6.0–14.5 years). The reference standard was based on histopathology with at least 3 years of clinical follow-up. The primary endpoint of the study was the mean area under the receiver operating characteristic curve (AUROC) for the AI system vs the radiologists in PDAC detection on CT. Noninferiority of AI was tested first with a margin of .05, with superiority tested if noninferiority was met.
Key Findings
Among all 3,440 patients, 1,103 received a diagnosis of pancreatic cancer.
In the sequestered testing cohort, pancreatic cancer was histologically confirmed in 406 of 1,130 patients. AI was associated with an AUROC of 0.92 (95% confidence interval [CI] = 0.90–0.93).
In the subset of 391 patients with CTs read by radiologists, 144 (37%) had histologically confirmed pancreatic cancer. AI was associated with statistically noninferior (P < .0001) and superior (P = .001) performance, with an AUROC of 0.92 (95% CI = 0.89–0.94), compared with an AUROC of 0.88 (95% CI = 0.85–0.91) for the pooled radiologists.
The investigators concluded: “AI demonstrated substantially improved pancreatic ductal adenocarcinoma detection on routine CT scans compared to radiologists on average, showing potential to detect cancer earlier and improve patient outcomes.”
Natalia Alves, MSc, of Diagnostic Image Analysis Group, Radboud University Medical Center, Nijmegen, the Netherlands, is the corresponding author for The Lancet Oncology article.
Disclosure: The study was funded by the European Union’s Horizon 2020 research and innovation programme. For full disclosures of all study authors, visit thelancet.com.
