543A machine-learning CMR approach to extract disease features and automate pulmonary arterial hypertension diagnosis

Abstract

Background: Pulmonary arterial hypertension (PAH) is a severe progressive condition. Quantitative cardiac magnetic resonance (CMR) imaging metrics target individual cardiac structures and have diagnostic and prognostic utility but are challenging to acquire. The aim of this study was to develop and test a tensor-based machine learning approach to holistically identify diagnostic features in PAH. Methods: Consecutive treatment naive patients with suspected PAH were identified. A tensor-based machine learning approach was developed and diagnostic accuracy compared to standard prospectively acquired CMR measurements using receiver operating characteristic analysis (ROC). Results: A cohort of 220 patients were identified: 150 with PAH and 70 with no pulmonary hypertension. The 150 with PAH includes 69 with idiopathic PAH (IPAH). The diagnostic accuracy of the developed machine learning approach was high as assessed by area under the curve at ROC (p < 0.001): 0.90 for PAH and 0.97 for IPAH, slightly higher than standard CMR metrics, see Table 1. Establishing the diagnosis was achieved within 10 seconds. Learnt features were visualised in feature maps with correspondence to cardiac phases, confirming known and also identifying new diagnostic features in PAH. The figure shows red learnt features of PAH (A and B) and green learnt features of normality (C and D) on patients with PAH and no PH, respectively. Conclusion: A tensor-based machine learning approach has been developed and applied to CMR. High diagnostic accuracy has been shown for PAH diagnosis and new learnt features were visualised with diagnostic potential.