Deep learning approaches for the detection of scar presence from cine cardiac magnetic resonance adding derived parametric images
2024

Using Deep Learning to Detect Heart Scars from MRI Images

Sample size: 206 publication 10 minutes Evidence: moderate

Author Information

Author(s): Francesca Righetti, Giulia Rubiu, Marco Penso, Sara Moccia, Maria L. Carerj, Mauro Pepi, Gianluca Pontone, Enrico G. Caiani

Primary Institution: Politecnico di Milano

Hypothesis

The use of parametric images derived from cine cardiac magnetic resonance images can improve the detection of myocardial scar tissue using deep learning.

Conclusion

The study demonstrates that a convolutional neural network can effectively classify the presence of myocardial scars in cardiac MRI images without the need for contrast agents.

Supporting Evidence

  • The CNN achieved an accuracy of 0.79 and an area under the ROC curve of 0.86 for slice classification.
  • At the patient level, the CNN exhibited a prediction accuracy of 1.0 for classifying patients as control or pathologic.
  • The study suggests that the proposed approach could serve as a preliminary screening tool for LGE-CMR imaging.

Takeaway

Researchers created a computer program that helps doctors find scars in the heart using special pictures from MRI scans, without needing any dye.

Methodology

The study used a convolutional neural network to analyze cine cardiac MRI images and parametric images derived from them to classify the presence of myocardial scars.

Potential Biases

Potential bias due to reliance on a single expert for ground truth labeling and the limited diversity of the patient population.

Limitations

The study had a relatively small dataset and relied on the interpretation of images by a single expert, which may affect the generalizability of the results.

Participant Demographics

The study included 206 patients, with 158 having ischemic dilated cardiomyopathy and 48 control patients.

Statistical Information

P-Value

p<0.0001

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.1007/s11517-024-03175-z

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication