Can artificial intelligence play a role in the analysis of non-gated, non-cardiac CT?

Jack Evans, Professor Joseph Suttie

Abstract

The rapid development of artificial intelligence (AI) has offered an opportunity to improve processing time and diagnostic accuracy of radiological images (1). Deep learning is a subfield of AI that can analyse unstructured data in its raw form, and determine features which distinguish different categories of data from one another, thereby minimising the need for human intervention (2). This is clearly very attractive for medicine, which deals with high volumes of raw data such as images and text. Cardiac imaging for risk stratification remains one of the most frequently performed medical investigations, reflecting the leading
role cardiovascular disease plays in mortality globally. The aim of this review is to consider the applications of AI and deep learning in cardiovascular imaging, in particular looking at assessment of coronary vessels on both gated and nongated CT chest.

Coronary artery calcium score is of most value when analysing asymptomatic patients with intermediate risk of coronary artery as assessed by the Framingham Risk Score (3). It is a semi-automated analysis of calcified plaques in the coronary arteries with density of >130 Hounsfield units (HU) on an ECG gated non contrast CT scan. Previous studies have used fully automated methods for assessing Coronary Artery Calcium score, where the AI algorithm identified calcification in 73.8% of cases, and assigns patients to the correct risk category in 93.4% of cases (4). Multiple other studies have looked at looked at more sophisticated subtypes of AI, including deep convolutional neural network (similar to that of deep learning) and found promising results in the assessment of ECG gated CT scans of the chest (5,6,7). AI has also been used to analyse dedicated CT scanning of the coronary arteries, including CTCA. A sub study of the CLARIFY trial (CT EvaLuation by Artificial Intelligence For Atherosclerosis, Stenosis and Vascular Morphology) evaluated the interobserver variability among expert readers for qualifying the volume of coronary plaque and plaque components on CTCA using an AI software as an index. The readers were consistent in identification of plaque volumes, however found a difference in the assessment of plaque composition (8).

The above examples look at dedicated imaging of the coronary arteries, however there are a significant number of non-dedicated CT chests performed each day for the evaluation of other structures within the thorax. The Society of Cardiovascular Computed Tomography and the Society of Thoracic Radiology
published a guideline in 2017 for coronary artery calcium scoring on noncontrast noncardiac CT scans and determined that coronary artery calcium scoring in all noncontrast chest examinations should be reported in those patients with moderate/severe coronary artery calcium undergoing lung cancer screening (9). Based on previous studies looking at dedicated coronary imaging, AI could also play a significant role in the assessment of coronary artery calcium in noncontrast CT scans. It’s limitations, within the early stages may be assessing the plaque composition, however as AI and deep learning mechanisms become more advanced, it may play a vital role in the assessment of coronary artery
calcium score in both research and clinical contexts.

References

1. Mintz, Y.; Brodie, R. Introduction to artificial intelligence in medicine. Minim.Invasive Ther.
   Allied Technol. 2019, 28, 73–81
2. Francesco Piccialli, Vittorio Di Somma, Fabio Giampaolo, Salvatore Cuomo, Giancarlo
   Fortino, A survey on deep learning in medicine: Why, how and when?, Information
   Fusion, Volume 66, 2021, Pages 111-137, ISSN 1566-2535, https://doi.org/10.1016/j.
   inffus.2020.09.006.
3. Greenland, P., LaBree, L., Azen, S. P., Doherty, T. M., & Detrano, R. C. (2004). Coronary
   artery calcium score combined with Framingham score for risk prediction in asymptomatic
   individuals. JAMA, 291(2), 210–215. https://doi.org/10.1001/jama.291.2.210
4. Isgum, I.; Rutten, A.; Prokop, M.; van Ginneken, B. Detection of coronary calcifications from
   computed tomography scans for automated risk assessment of coronary artery disease. Med.
   Phys. 2007, 34, 1450–1461.
5. Takx, R.A.; de Jong, P.A.; Leiner, T.; Oudkerk, M.; de Koning, H.J.; Mol, C.P.; Viergever,
   M.A.; Isgum, I. Automated coronary artery calcification scoring in non-gated chest CT:
   Agreement and reliability. PLoS ONE 2014, 9, e91239.
6. Cano-Espinosa, C.; Gonzalez, G.; Washko, G.R.; Cazorla, M.; Estepar, R.S.J. Automated
   Agatston Score Computation in non-ECG Gated CT Scans Using Deep Learning. Proc.
   SPIE Int. Soc. Opt. Eng. 2018, 10574, 673–678.
7. Sandstedt, M.; Henriksson, L.; Janzon, M.; Nyberg, G.; Engvall, J.; De Geer, J.; Alfredsson,
   J.; Persson, A. Evaluation of an AI-based, automatic coronary artery calcium scoring software. Eur. Radiol. 2020, 30, 1671–1678.
8. Jonas, R.A.; Weerakoon, S.; Fisher, R.; Griffin, W.F.; Kumar, V.; Rahban, H.; Marques, H.;
   Karlsberg, R.P.; Jennings, R.S.; Crabtree, T.R.; et al. Interobserver variability among expert
   readers quantifying plaque volume and plaque characteristics on coronary CT angiography:
   A CLARIFY trial sub-study. Clin. Imaging 2022, 91, 19–25
9. Hecht, H.S.; Cronin, P.; Blaha, M.J.; Budoff, M.J.; Kazerooni, E.A.; Narula, J.; Yankelevitz,
   D.; Abbara, S. 2016 SCCT/STR guidelines for coronary artery calcium scoring of
   noncontrast noncardiac chest CT scans: A report of the Society of Cardiovascular Computed
   Tomography and Society of Thoracic Radiology. J. Thorac. Imaging 2017, 32, W54–W66.