Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
March 6, 2026European Heart Journal - Digital HealthOpen Access

Multimodal deep learning predicts 30-day MACE with AUROC ~0.82, outperforming RCRI and human CAD-RADS.

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Major adverse cardiac events significantly impact perioperative morbidity and mortality, creating a need to optimize risk prediction.

Does a fully automated multimodal deep learning system integrating CCTA and clinical data improve 30-day MACE prediction compared to RCRI and CAD-RADS alone in patients undergoing elective non-cardiac surgery?

Population

639 patients undergoing CCTA for perioperative risk assessment before elective non-cardiac surgery

Comparison

Multimodal DL system vs RCRI and CAD-RADS alone

Design

Model development and validation study

Follow-up

30 days

Key result

A multimodal deep learning system integrating automated CAD-RADS, anatomy, and demographics predicted 30-day MACE with AUROC 0.82, outperforming RCRI and human CAD-RADS.

Authors

JLJuan LuGHG. HuangfuAIAbdul Rahman Ihdayhid

Discussion

Loading...

Member takes

Overview

May improve MACE prediction before elective noncardiac surgery; extends RCRI and CAD-RADS but should not yet change practice.

Key Points

  • The research aims to develop a deep learning system to improve the prediction of major adverse cardiac events (MACE) in high-risk perioperative patients.
  • Included 639 patients undergoing CCTA for risk assessment.
  • Used convolutional neural networks to identify CAD-RADS scores and segment cardiac structures.
  • Combined imaging features with demographics and comorbidities to predict MACE risk.
  • Evaluated performance using AUROC curves against the revised cardiac risk index.
  • 45 patients experienced MACE within 30 days.
  • Multimodal deep learning system achieved AUROC of 0.82, outperforming CAD-RADS (0.69) and RCRI.
  • System demonstrated sensitivity of 83% and specificity of 79%.

Structured PICO

Does a fully automated multimodal deep learning system integrating CCTA and clinical data improve 30-day MACE prediction compared to RCRI and CAD-RADS alone in patients undergoing elective non-cardiac surgery?

P
Population
639 patients undergoing coronary computed tomography angiography (CCTA) as part of perioperative risk assessment for elective non-cardiac surgery (61% orthopaedic, 27% vascular, remainder abdominal/pelvic or spine), mean age 70±9 years, 56% males.
I
Intervention
Fully automated multimodal deep learning (DL) system integrating patient demographics, comorbidities, and CCTA findings (automated CAD-RADS scores and segmentations of the left ventricle, aorta, and heart).
C
Comparator
Revised cardiac risk index (RCRI), human expert CAD-RADS analysis, and automated CAD-RADS alone.
O
Outcome
Major adverse cardiac events (MACE) within 30 days.hard clinical

A fully automated deep learning system combining CCTA imaging features with clinical data significantly improves 30-day perioperative MACE prediction compared to standard clinical risk scores.

Cite This Study

Lu et al. (2026) studied this question. A multimodal deep learning system integrating automated CAD-RADS, anatomy, and demographics predicted 30-day MACE with AUROC 0.82, outperforming RCRI and human CAD-RADS.

synapsesocial.com/papers/69aa7087531e4c4a9ff5a6behttps://doi.org/10.1093/ehjdh/ztag037
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Clinical Validation of a Deep Learning Algorithm for Automated Coronary Artery Disease Detection and Classification Using a Heterogeneous Multivendor Coronary Computed Tomography Angiography Data Set2024 · 5 citations
  2. 2Multimodal deep learning to predict postoperative major adverse cardiac and cerebrovascular events after noncardiac surgery2025 · 6 citations
  3. 33D Convolutional Neural Network for Predicting Clinical Outcome from Coronary Computed Tomography Angiography in Patients with Suspected Coronary Artery Disease2025
  4. 4Comparison of conventional scoring systems to machine learning models for the prediction of major adverse cardiovascular events in patients undergoing coronary computed tomography angiography2022 · 5 citations
  5. 5The prognostic value of automated coronary calcium derived by a deep learning approach on non-ECG gated CT images from 82Rb-PET/CT myocardial perfusion imaging2021 · 19 citations