Opportunistic Assessment of Ischemic Heart Disease Risk Using Abdominopelvic Computed Tomography and Medical Record Data: a Multimodal Explainable Artificial Intelligence Approach

Abstract

Current risk scores for predicting ischemic heart disease (IHD) risk—the leading cause of global mortality—have limited efficacy. While body composition (BC) imaging biomarkers derived from abdominopelvic computed tomography (CT) correlate with IHD risk, they are impractical to measure manually. Here, in a retrospective cohort of 8,197 contrast-enhanced abdominopelvic CT examinations undergoing up to 5 years of follow-up, we developed improved multimodal opportunistic risk assessment models for IHD by automatically extracting BC features from abdominal CT images and integrating these with features from each patient’s electronic medical record (EMR). Our predictive methods match and, in some cases, outperform clinical risk scores currently used in IHD risk assessment. We provide clinical interpretability of our model using a new method of determining tissue-level contributions from CT along with weightings of EMR features contributing to IHD risk. We conclude that such a multimodal approach, which automatically integrates BC biomarkers and EMR data can enhance IHD risk assessment and aid primary prevention efforts for IHD.

Use of Multi-Modal Data and Machine Learning to Improve Cardiovascular Disease Care

“…However, the performance of these scores, measured by the area under the receiver operating characteristic curve (AUC), has been modest when testing them in more diverse patient populations. Thus, Chaves et al developed a framework to use deep learning and machine learning models that enable opportunistic risk assessment for ischemic heart disease (IHD) using automatically measured imaging features from abdominopelvic CT examinations in combination with information from the patient's EMR ( 15 ).…”

“… Architecture of multi-modal data fusion combining Imaging and clinical data. Figure taken from Chaves et al ( 15 ). In their described framework, readily available CT images are combined with clinical data (e.g.…”

“…Another approach is to first extract some features and measurements from each data modality and then combine this subset of features. As an example, in Chaves et al to construct the Segmentation + Clinical fusion model, average muscle radiodensity and the VAT/SAT ratio were first extracted and then combined with clinical data ( 15 ). In contrast, for late fusion, or decision level fusion, first, for every single modality a model is trained [for example, Zhao et al' EHR and genetic late fusion model ( 19 )].…”

Fusion Modeling: Combining Clinical and Imaging Data to Advance Cardiac Care

“…These models, while innovative, fall short of fusing imaging data directly with clinical information. Chaves et al 38 fused L3 slices from CT for body composition analysis, and known clinical risk factors through late fusion, for opportunistic screening for ischemic heart disease. They observed that fusion modeling outperforms single-modality models in terms of screening performance.…”

Cardiovascular disease and all-cause mortality risk prediction from abdominal CT using deep learning

“…In this work we focus on risk scoring from abdominal CT scans, which has only been studied previously in a handful of works. 9, 27, 28 Abdominal CT are rich with biomarkers known to be relevant to cardiovascular risk such as visceral fat and aortic plaque. 29, 30 This is the first work to feed the entire abdominal CT scan into a deep learning model.…”