Related Experiment Video
Updated: Aug 22, 2026
High-Resolution Cardiac Positron Emission Tomography/Computed Tomography for Small Animals
Published on: December 16, 2022
PULSE: A Unified Multi-Task Architecture for Cardiac Segmentation, Diagnosis, and Few-Shot Cross-Modality Clinical
Insights
PULSE is a unified framework for cardiac image analysis, integrating segmentation, diagnosis, and reporting. This vision-based system achieves high accuracy in ventricular segmentation and cardiomyopathy diagnosis, streamlining clinical workflows.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Cardiology
Background:
- Accurate cardiac image analysis is crucial for diagnosing heart conditions.
- Current methods often use separate models for segmentation, classification, and reporting, hindering clinical application.
- A unified approach can improve efficiency and accuracy in cardiac diagnostics.
Purpose of the Study:
- To introduce PULSE, a novel, unified framework for cardiac image analysis.
- To integrate ventricular segmentation, cardiomyopathy diagnosis, and structured clinical reporting into a single system.
- To demonstrate the framework's effectiveness and efficiency in clinical settings.
Main Methods:
- PULSE utilizes a Vision Transformer (ViT-B/14) with a Dense Prediction Transformer (DPT) decoder for segmentation.
- Cardiomyopathy diagnosis is performed using a Random Forest classifier on clinical biomarker data.
- A template-based module generates structured clinical reports with automated flags and indices.
Main Results:
- Achieved a mean Dice score of 88.8% for ventricular segmentation on the ACDC benchmark.
- Reached 90.0% patient-level diagnostic accuracy for cardiomyopathy with a macro-AUC of 0.982.
- Generated reports showed 92.7% agreement on clinical flags, with LVEF MAE of 3.09%.
Conclusions:
- PULSE offers a unified, vision-based solution for multiple cardiac image analysis tasks.
- The framework demonstrates high performance across segmentation, diagnosis, and reporting.
- PULSE shows potential for improved clinical deployment and cross-modality transfer learning.
Abstract:
Cardiac image analysis requires accurate ventricular segmentation, disease classification, and structured clinical reporting; these tasks are typically handled by separate models, limiting clinical deployment. To address this, we introduce PULSE, a unified three-task framework that performs: (1) ventricular segmentation using a DINOv2 Vision Transformer (ViT-B/14) backbone with a Dense Prediction Transformer (DPT) decoder and deep supervision; (2) cardiomyopathy diagnosis via a 23-dimensional clinical biomarker vector fed to a Random Forest classifier; and (3) a structured, template-based clinical reporting module that populates a predefined report with the measured indices and rule-based abnormality flags. The framework is entirely vision-based: all reported text is produced by deterministic, rule-driven templates. Additionally, the model takes 2.5D inputs (three adjacent short-axis slices) and is evaluated via a 5-fold stratified ensemble. With extensive experiments on the ACDC benchmark, PULSE achieves a mean Dice of 88.8% (RV: 90.3%, Myo: 84.7%, LV: 91.6%, HD95 $\leq$ 4.6 mm), 90.0% patient-level diagnostic accuracy (macro-AUC 0.982), and 92.7% clinical flag agreement in the generated reports (LVEF MAE 3.09%, within inter-observer tolerance). Without retraining, PULSE achieves 85.3% mean Dice on M&Ms-2 (360 subjects, RV: 87.9%, Myo: 80.4%, LV: 87.5%) and 88.1% LV Dice on Sunnybrook MRI. We further demonstrate that few-shot fine-tuning on CAMUS echocardiography samples yields a mean Dice of 73.2%, showing strong cross-modality transfer from cardiac MRI priors.
