At the 2024 CCA Summit, Iwan Paolucci, PhD, presented on the topic “Innovations of Artificial Intelligence in Cholangiocarcinoma” in which he described the many applications of artificial intelligence (AI) in oncology, including drug discovery, diagnosis, prognosis, and treatment.1 One example of AI in drug discovery that Dr Paolucci mentioned was the development of an AI algorithm by Charan et al, which analyzed a data set of 2356 chemical compounds to identify those that are effective in inhibiting FGFRs.2 Four machine-learning algorithms were used to train prediction models based on chemical descriptors; the random forest-based prediction model achieved the highest accuracy on the training (98.9%), test (89.8%), and external test (90.3%) data sets. A further illustration of AI in drug discovery is the initiative by Google referred to as AlphaFold, which utilizes an AI algorithm to anticipate the folding structure of a protein derived from its sequence.1 These algorithms can be used to screen compounds for specific properties, which can lead to discovery of new drugs.
AI can be used to discover new drugs; it can also be used to diagnose disease. Dr Paolucci described one example in which a group of researchers built a radiomics model using magnetic resonance imaging, as well as serum carcinoembryonic antigen level and tumor diameter, to distinguish between intrahepatic mass-forming cholangiocarcinoma (CCA) and colorectal liver metastasis.3 Radiomics involves the extraction, selection, and analysis of quantitative data from images that cannot be discerned through visual examination. This process aims to represent tumor pathophysiology and intratumor heterogeneity, ultimately enhancing clinical decision-making. This model is a novel noninvasive tool that can be used to help guide treatment and prognosis.3
In addition, AI can be used to predict early recurrence of cancer. A multicenter study analyzed 311 patients with intrahepatic CCA who underwent resection and used AI-based computed tomography (CT) radiomics to predict if they would have recurrence of disease.4 This combined clinical-radiomics model included 15 radiomic features and 3 clinical features to predict prognosis. Dr Paolucci showed 2 patient scans that the model was able to distinguish; 1 patient was determined to have a low risk for recurrence, with a recurrence-free survival of 92 weeks, and the other patient was deemed as high risk for recurrence, with a recurrence-free survival of 34 weeks.1
AI can also be used to aid in the treatment of disease. One study used AI to segment 104 anatomical structures, including blood vessels and skeletal structures in CT images, a process that only takes approximately 2 to 3 minutes. This process is useful because it helps to improve planning for radiotherapy, surgery, thermal ablation, and embolization.1
To summarize, AI can be utilized in all stages of CCA research. AI has significant potential in drug discovery to identify promising compounds, and radiomics shows promise in diagnosis and response tracking. Overall, AI is a powerful tool for rapid-image processing that can influence many downstream applications.1
To sign up for our newsletter or print publications, please enter your contact information below.
