Multi-channel deep learning model-based myocardial spatial–temporal morphology feature on cardiac MRI cine images diagnoses the cause of LVH

Insights into Imaging

Table 2 Three-class diagnostic performance of the Model 3 and radiologists/cardiologists in the external test dataset

	Three-class Cohen’s kappa	Three-class overall accuracy	p value (vs model 3)	Reader agreement (vs model 3)
DL model 3	0.693	77.4% (41/53)	Reference	Reference
Expert cardiovascular radiologist	0.717	79.2% (42/53)	0.814	0.552
Senior radiologist	0.582	67.9% (36/53)	0.276	0.502
Junior radiologist 1	0.355	45.3% (24/53)	< 0.001	0.162
Junior radiologist 2	0.340	43.4% (23/53)	< 0.001	0.163
Senior cardiologist	0.478	58.5% (31/53)	0.037	0.134
Junior cardiologist 1	0.405	50.9% (27/53)	0.005	0.147
Junior cardiologist 2	0.352	45.3% (24/53)	< 0.001	0.195

Italic indicates the results of DL model 3 as the reference, and report if the performance of any other rater (radiologists or cardiologists) had any difference from DL model 3