AI-derived cardiovascular border parameters from preoperative chest radiographs were independently associated with postoperative in-hospital MACE and provided modest incremental predictive information beyond routinely available clinical factors, though further external validation is required.
Postoperative in-hospital MACE occurred in 1.6% of a large surgical cohort.
All ADC-derived cardiomediastinal vascular border (CVB) parameters remained independently associated with MACE after multivariable adjustment.
Cardiothoracic (CT) ratio was independently associated with postoperative MACE and had the highest standalone AUC reported.
The composite CVB z-score was independently associated with postoperative MACE.
Adding the CT ratio or composite CVB z-score to a clinical reference model modestly improved discrimination.
The incremental improvement was similar in temporal validation.
Automated CVB analysis may extract additional cardiovascular risk information from routine preoperative CXR without additional imaging or patient burden.
This study looked at whether an artificial intelligence tool could get more useful information out of the chest X-rays that many patients already have before surgery. The tool measures the size and outline of the heart and major blood vessels. The researchers analyzed records from over 101,000 patients who had surgery under general anesthesia at one large hospital, and about 1.6% of them had a major cardiovascular event (such as a serious heart complication) while in the hospital after surgery. The AI-derived measurements, including the ratio of heart size to chest width and a combined score of vessel border measurements, were each linked to a higher risk of these events, even after accounting for other factors such as age and health conditions. When the researchers added these measurements to a prediction model based on regular clinical information, the model's ability to distinguish patients who would and would not have an event improved modestly (from 0.894 to 0.906 on a standard 0–1 scale). A similar improvement was seen when testing on a later time period. This research suggests that routine pre-surgery chest X-rays might offer extra risk information without additional tests or burden on patients. However, the gain was small, the study was done at a single center and looked back at existing records, and it has not been tested at other hospitals. The authors note that further external validation is needed before this approach can be considered clinically useful.
Jang H, Kim J, Oh H, Yoon H, Kim C, Kim W, et al.. (2026). AI-Powered Preoperative Chest Radiograph-Derived Cardiovascular Border Indices and Risk of Postoperative Major Adverse Cardiovascular Events: A Large-Scale Retrospective Cohort Study.. Medicina (Kaunas, Lithuania). https://doi.org/10.3390/medicina62101897