CMR-CLIP

Cleveland Clinic / Case Western (Nakashima et al.)

Cardiac MRI

Filter catalog by Modality:
Cardiac MRI

Clinical text

Filter catalog by Modality:
Text & EHR

Multimodal

Filter catalog by Modality:
Multimodal

General Purpose / Multi-task

Filter catalog by Disease / Trait:
General / Foundation

Non-ischemic cardiomyopathy

Filter catalog by Disease / Trait:
Structural Heart & Cardiomyopathy

Ischemic cardiomyopathy

Filter catalog by Disease / Trait:
Coronary & Ischemic Disease

Cardiac amyloidosis

Filter catalog by Disease / Trait:
Structural Heart & Cardiomyopathy

LV systolic dysfunction (LVSD)

Filter catalog by Disease / Trait:
Cardiac Function & Hemodynamics

LV dilation

Filter catalog by Disease / Trait:
Structural Heart & Cardiomyopathy

Left ventricular hypertrophy (LVH)

Filter catalog by Disease / Trait:
Structural Heart & Cardiomyopathy

Multi-label classification

Filter catalog by Task Type:
Classification

Binary classification

Filter catalog by Task Type:
Classification

Embedding

Filter catalog by Task Type:
Representation Learning

Hybrid

Filter catalog by Architecture:
Hybrid / Multi-branch

PyTorch

Filter catalog by Framework:
PyTorch

MIT

Filter catalog by License:
Permissive

file_copy

View model files code View code

Vision-language model that jointly embeds a cardiac MRI study, treated as video, with the impression section of its clinical report. Combines a video encoder over cine/LGE frame sequences with a Bio+ClinicalBERT text encoder using CLIP-style contrastive training. Supports zero-shot and few-shot classification of cardiomyopathies, amyloidosis, and LV dysfunction, plus image/report retrieval and structured report drafting. Trained on a private, single-institution corpus of roughly 11,000-14,000 CMR study-report pairs from Cleveland Clinic and Case Western.

memory Specifications

gavel License

MIT

check_small Open source check_small Commercial use OK check_small Modifications allowed Attribution required No share-alike requirement

License for model weights only. Associated code may be licensed seperately, check code source for specific terms.

description Publication

Contrastive language image pretraining for a cardiac magnetic resonance image embedding with zero-shot capabilities open_in_new

2026 · original paper

DOI: 10.1038/s41467-026-73022-2

database Training & evaluation data

Single-Institution CMR Study-Report Corpus (CMR-CLIP)

train

public 12,500 subjects

12,500 patients (14,214 paired CMR study-report pairs used for training; 41,936 studies pre-exclusion); Cleveland Clinic, Ohio

science Capabilities & performance

Disease/function screening (avg AUC)

Multi-label classification General Purpose / Multi-task

0.764 AUROC Internal held-out test set · internal

0.737 AUROC External geographically distinct dataset · external

Non-ischemic cardiomyopathy (zero-shot)

Binary classification Non-ischemic cardiomyopathy

0.885 Accuracy Internal held-out test set · internal

Ischemic cardiomyopathy (zero-shot)

Binary classification Ischemic cardiomyopathy

0.88 Accuracy Internal held-out test set · internal

Cardiac amyloidosis (zero-shot)

Binary classification Cardiac amyloidosis

0.962 Accuracy Internal held-out test set · internal

LV systolic dysfunction (zero-shot)

Binary classification LV systolic dysfunction (LVSD)

0.86 AUROC Internal held-out test set · internal

LV dilation (zero-shot)

Binary classification LV dilation

0.79 AUROC Internal held-out test set · internal

LV hypertrophy (zero-shot)

Binary classification Left ventricular hypertrophy (LVH)

0.82 AUROC Internal held-out test set · internal

Embedding

General Purpose / Multi-task

CVAI Catalog

View Catalog

CVAI Catalog

CMR-CLIP

memory Specifications

gavel License

description Publication

database Training & evaluation data

science Capabilities & performance