Deep Learning Replaces Visual Analysis by Skilled Operators for Cardiac Amyloidosis Diagnosis by Cine-CMR


This article was originally published here

Diagnosis (Basel). 2021 Dec 29;12(1):69. doi: 10.3390/diagnostics12010069.


BACKGROUND: The diagnosis of cardiac amyloidosis (CA) from cine-CMR (cardiac magnetic resonance) alone is unreliable. In this study, we tested whether a convolutional neural network (CNN) could outperform the visual diagnosis of experienced operators.

Method: 119 patients with cardiac amyloidosis and 122 patients with left ventricular hypertrophy (LVH) of other origins were retrospectively selected. Diastolic and systolic cine-CMR images were preprocessed and labeled. A double-entry visual geometry group (VGG) model was used for binary image classification. All images belonging to the same patient were distributed in the same set. Precision and area under the curve (AUC) were calculated per image and per patient from a set of tests retained at 40%. The results were compared with a visual analysis evaluated by three experienced operators.

RESULTS: Frame-based comparisons between humans and a CNN provided an accuracy of 0.605 versus 0.746 (p p p p

CONCLUSION: on the basis of cine-CMR images alone, a CNN is able to discriminate cardiac amyloidosis from LVH of other origins better than experienced human operators (15 to 20 points more in absolute value for accuracy and AUC), demonstrating a unique ability to identify what the eyes cannot see through conventional X-ray analysis.

PMID:35054236 | DOI:10.3390/diagnostics12010069


Comments are closed.