AI pathology model assists in liver cancer diagnosis

A confidence-based artificial intelligence (AI) biomarker demonstrates clinical utility in the diagnosis of liver cancer, according to a real-world study.

“By accurately discriminating intrahepatic cholangiocarcinoma (ICCA) from metastasis, this tool offers the potential to reduce unnecessary investigations and accelerate therapeutic decision-making,” the authors said.

A retrospective analysis was performed on 544 patients across five European centres, including cases of either ICCA or metastases from extrahepatic cancers. Three deep-learning architectures using foundation models were examined, namely CTranspath/HistoBistro, UNI/CLAM, and CONCH/TITAN.

The authors assessed the performance of these AI tools using the area under the receiver operating characteristic curve (AUROC) and the false-positive rate (FPR). They also implemented a confidence estimation system using the generalized ODIN (G-ODIN) approach, utilizing predictive entropy as metric.

Lastly, the authors prospectively validated the final model, AI²CCA, in 161 patients across four international centres in France, India, and Korea.

The CONCH/TITAN architecture showed the best performance (AUROC 0.840) in the retrospective test set. Evidently greater predictive entropy derived via G-ODIN was seen in misclassified cases, validating its use as a confidence metric. Confidence thresholding also improved the AUROC to 0.958, with an FPR of 0, while retaining 46 percent of samples for high-confidence prediction.

Prospective validation showed AI²CCA attaining an AUROC of 1.00 in the French cohort and 0.965 in the Asian cohorts, with only one misclassified case in the Asian series.

“ICCA is a rare but highly lethal adenocarcinoma arising within the hepatic parenchyma. Diagnosis presents a significant clinical challenge as the histological features of ICCA substantially overlap with those of metastatic liver cancers,” the authors said.

“This diagnostic ambiguity often necessitates extensive and costly exclusionary investigations, such as upper and lower gastrointestinal endoscopy, to rule out an occult primary site. [T]his process results in treatment delays and an increased financial burden on health care systems,” they added.