Automated identification of cardiac amyloidosis using cross-modal neural networks on [Formula: see…

AI / informaticsThe international journal of cardiovascular imaging· Jul 2026

Automated identification of cardiac amyloidosis using cross-modal neural networks on [Formula: see text]-Pyrophosphate SPECT imaging and clinical data.

Fatih Batı, Nilüfer Bıçakcı, Musa Aydın … Berna Kiraz

Cardiovascular ImagingAI / informaticsDiagnosisQ2IF 1.7

Fusion AI model combining scan imaging with clinical data detects cardiac amyloidosis with 90.9% accuracy.

Automated identification of cardiac amyloidosis using cross-modal neural networks on [Formula: see text]-Pyrophosphate SPECT imaging and clinical data.

Fatih Batı … Berna Kiraz

The international journal of cardiovascular imaging · Jul 2026 · Q2 in Radiology

Background

Accurate, early identification of transthyretin cardiac amyloidosis (ATTR-CA) is challenging yet critical for effective treatment.

Methods

On a curated cohort of 109 patients (62 positive, 47 negative), fusion models consistently outperformed image-only convolutional neural networks (convolutional neural networks): Cross-Modal Fusion Network raised average accuracy by 6.9 percentage points and Late Fusion by 5.4.

n = 109 patients

90.9%

Results

90.9%

fusion of imaging and clinical data correctly identified amyloidosis most of the time

n = 109 patients

More results

Notable gains included a +30.8 percentage points sensitivity(recall) for ResNet-34 with Cross-Modal Fusion Network and ResNet-50 Late Fusion sensitivity of 94.9%.

“

Conclusion

These results show that integrating imaging and text/numeric clinical data yields superior, reproducible detection of [Formula: see text] scan positivity and may streamline scan interpretation.

Automated identification of cardiac amyloidosis using cross-modal neural networks on [Formula: see text]-Pyrophosphate SPECT imaging and clinical data.

Automated identification of cardiac amyloidosis using cross-modal neural networks on [Formula: see text]-Pyrophosphate SPECT imaging and clinical data.

Related papers