Published
The 80th Korean Congress of Radiology (KCR 2024)
Authors
Heejun Shin1, Taehee Kim1, Dongmyung Shin1, Joshua Ra2
Affiliations
1Radisen, Korea, Republic of
2UCSF Medical Center, USA
shinsae11@radisentech.com
*to be published soon.
Purpose
This study aimed to evaluate a commercially available AI model that measures the cardiothoracic ratio (CTR) on chest x-ray radiographs (CXRs) and examines its association with echocardiographic diagnoses of severe left ventricular hypertrophy (SLVH) and dilated left ventricle (DLV).
Materials and Method
The study utilized 71,589 CXRs from 24,689 patients and employed the CheXchoNet dataset. Patients underwent echocardiograms within one year of their X-ray scans, which diagnosed them with either SLVH or DLV. A binary label was created based on the prese nce of either condition (9,861/61,728 composite positive/negative). Commercially available AI software (AXIR-CX; Radisen Co. Ltd.) measured CTRs on individual CXRs. Statistical analyses included examining CTR value distributions according to the composite labels using a t-test and calculating sensitivity, specificity, area under the curve (AUC), and the Youden index with ground truth to the composite labels. We also developed a separate classification AI (a multilayer perceptron with two fully connected layers) that took the CTR value, age, and sex of a patient as inputs to predict the binary composite label. 80% of CXRs were used for training and the remaining for evaluation.
Results
The average CTR value was significantly higher in composite SLVH/DLV positive patients compared to negative patients (mean ± std: 0.56 ± 0.07 for positive; 0.51 ± 0.07 for negatives; p-value < 0.001). The binary classification of the composite label showed an AU C of 0.69. The Youden index of 0.30 was maximized when CTR > 0.53, reporting a sensitivity of 0.70 and specificity of 0.60. When sex, age, and CTR were used as inputs in the classification AI model, an AUC was 0.71, and the Youden index was 0.32, with a sensitivity of 0.74 and specificity of 0.58. The exclusion of CTR values as inputs significantly dropped the AUC to 0.54, highlighting the importance of CTR measurements for accurately predicting composite SLVH/DLV labels.