Cholangiocarcinoma Foundation logo
DONATE

Decoding Artificial Intelligence: How Smart Technology Can Spot What Clinicians Might Miss

Yashbir Singh, ME, PhD1, Jesper B Andersen, MSc, PhD2, Gregory J Gores, MD3

Affiliations: 
1Radiology, Mayo Clinic, Rochester, Minnesota, USA
2Biotech Research and Innovation Centre (BRIC), Department of Health and Medical Sciences, University of Copenhagen, Denmark
3Division of Gastroenterology & Hepatology, Mayo Clinic, Rochester, Minnesota

Artificial Intelligence in Clinical Management

You go to your physician with a continuous unexplained issue – maybe nothing as obvious as yellowing pigmentation (jaundice) or continuous itching (pruritus) of your skin, but, for example, a longer-term abdominal or severe lower back pain. 

blood flow alterations, enlargedAs part of your examination, you are referred to imaging. When clinicians look at your scans for disease processes, they search for specific signs – abnormal tissue density, alterations in blood flow, an enlargement of structures such as bile ducts, or suspicious masses. But what if there are subtle changes that even experienced eyes often cannot detect them? This is where artificial intelligence (AI) becomes your powerful ally – a tireless friend that may support the clinician in an earlier diagnosis. 

AI uses something called “convolutional neural networks – CNN” – think of CNNs as a digital brain with millions of connections that learn by example in a repetitive manner – many examples are necessary to train this “computer brain”. When we feed it digital images obtained from CT or MRI scans, it breaks down each image into tiny squares called pixels, analyzing brightness, contrast, and patterns in ways human eyes simply cannot. The AI driven neural networks can AI-driven see minute details and recognize patterns in these images. 

Here is the fascinating part: cholangiocarcinoma often starts with microscopic changes in cell arrangement in the biliary tree that alter how this tissue area reflects imaging signals. While a radiologist sees the overall picture, AI can detect when just 2% of pixels show abnormal patterns – very small changes representing early cellular disruption that will not be visible to the experienced human eye for months or sometimes even years. The science behind this is called “radiomics” – extracting hundreds of mathematical features from your scans. AI measures things like:

  • Texture heterogeneity: How uniform or mixed the tissue appears at microscopic levels
  • Signal intensity variations: Tiny differences in how a tissue absorbs the contrast dye
  • Morphological signatures: Mathematical descriptions of shape irregularities that predict tumor behavior

Think of it like analyzing a photograph. You see a face (one image), but AI sees 10,000 data points – the exact curve of each feature, subtle color variations, and symmetry measurements. In cholangiocarcinoma detection, these “data points” might reveal that bile duct walls are 0.3mm thicker than normal or that tissue density has changed by just 3% – alterations too small for human detection but significant enough to signal a possible early cancer and warrant further analysis. 

Importantly, AI also performs “temporal analysis” – comparing your scans over time with mathematical precision. AI will detect if a suspicious area grew by just 1 millimeter or if tissue density increased by 5 Hounsfield units (a tissue density measurement used in CT scans). These precise measurements help your clinician understand not just if cancer is present, but how aggressive it might be.

The breakthrough is that AI learns from outcomes. When it analyzes a scan later confirmed as early cholangiocarcinoma, it updates its understanding, becoming more accurate with each case. This continuous learning means AI gets better at detecting your specific type of cholangiocarcinoma, potentially catching it months or years earlier than traditional methods.

References:

  1. Lambin, P., Rios-Velazquez, E., Leijenaar, R., Carvalho, S., Van Stiphout, R. G., Granton, P., … & Aerts, H. J. (2012). Radiomics: extracting more information from medical images using advanced feature analysis. European journal of cancer, 48(4), 441-446.
  2. Gillies, R. J., Kinahan, P. E., & Hricak, H. (2016). Radiomics: images are more than pictures, they are data. Radiology, 278(2), 563-577.
  3. Yamamoto, K., Hirakawa, N., & Itoi, T. (2022). Acute cholecystitis associated with a novel integrated biliary stent and nasobiliary drainage catheter system. Journal of Hepato-biliary-pancreatic Sciences, 30(5), e28-e30.
  4. Singh, Y., Jons, W. A., Eaton, J. E., Vesterhus, M., Karlsen, T., Bjoerk, I., … & Erickson, B. J. (2022). Algebraic topology-based machine learning using MRI predicts outcomes in primary sclerosing cholangitis. European radiology experimental, 6(1), 58.