Is There a Role For Artificial Intelligence in Cholangiocarcinoma Diagnosis?

Basil Njei, MD, MPH, PhD, Yale University, New Haven, CT, discusses his research into the use of artificial intelligence in the detection of malignant biliary strictures and cholangiocarcinoma.

Dr Njei explained the results found increasing evidence to support the use of artificial intelligence. The most promising method was convolutional neural network-based machine learning of cholangioscopy images, with convolutional neural network with endoscopic ultrasound demonstrating the best clinical performance application.

Transcript:

Hi, I'm Dr. Basil Njei. I am a faculty in the section of digestive diseases at Yale University, and I'm here to present to you the findings of our recent publication in the Annals of Gastroenterology. The title of our publication is “Artificial Intelligence in Endoscopic Imaging for Detection of Malignant Biliary Strictures and Cholangiocarcinoma: A Systematic Review.” This work was done in collaboration with doctors at Harvard, Yale, Oxford, Johns Hopkins, University of Utah, and the Methodist Hospital, as well as the Digestive Health Institute in Orlando.

A little bit of the background of cholangiocarcinomas: Cholangiocarcinoma is a very deadly cancer. Every year up to 8 000 people are diagnosed with bile duct cancer in America. This bile duct cancer or cholangiocarcinoma is so deadly that less than 35% of people survive, even after they receive curative treatment, after 5 years. Therefore, there is a need to diagnose this cancer early so as to treat it in time to prevent complications.

Unfortunately, most of the available techniques for diagnosis have yielded really poor results. Endoscopic retrograde cholangiopancreatography, which is a technique used to diagnose this disease by going through the bile ducts, has resulted in really low sensitivity. As a matter of fact, the most used technique in diagnosis of this cancer, cholangioscopy, only yields a sensitivity of about 65%, meaning that we have the possibility of missing about 35% of these cases. However, artificial intelligence when applied to computer-vision using convolutional neural networks is a promising tool in the diagnosis of this disease.

We carried out a systematic review to summarize available data on the diagnostic utility of endoscopic artificial intelligence-based techniques in diagnosing cholangiocarcinoma. Our study included 5 major studies put together, and we put together up to 1 500 patients in this study. We found that artificial intelligence when combined to cholangioscopy improved the sensitivity from about 65% to 94.7% and also increases specificity to 92%. Therefore, adding artificial intelligence resulted in an accuracy of up to 94.9% for diagnosing cholangiocarcinoma.

Our results suggest that there is increasing evidence to support the role of artificial intelligence in diagnosis of malignant biliary strictures and cholangiocarcinoma. Convolutional neural networks, which is a specific type of artificial intelligence, based on machine learning of cholangioscopy images appears to be the most promising, while convolutional neural networks with endoscopic ultrasound seems to have the best performance in application.

In the future, we expect that this artificial intelligence technique will be attached to regular endoscopic devices so that we can diagnose patients in real-time. Hopefully, these techniques can be adopted in real life in many clinics and around the world to improve diagnosis of this deadly disease. Thank you.

Source:

Njei B, McCarty TR, Mohan BP, Fozo L, and Navaneethan U. Artificial intelligence in endoscopic imagining for detection of malignant biliary strictures and cholangiocarcinoma: A systemic review. Ann Gastroenterol. 2023; 36(2):223-230. doi:1.20524/aog.2023.0779