Background: Doppler echocardiography analysis has become a golden standard in the modern diagnosis of heart diseases. In this paper, we propose a set of techniques for semi-automated parameter extraction for aortic valve stenosis severity grading.
Objectives: The main objectives of the study is to create echocardiography image processing techniques, which minimize manual image processing work of clinicians and leads to reduced human error rates.
Methods: Aortic valve and left ventricle output tract spectrogram images have been processed and analyzed. A novel method was developed to trace systoles and to extract diagnostic relevant features. The results of the introduced method have been compared to the findings of the participating cardiologists.
Results: The experimental results showed the accuracy of the proposed method is comparable to the manual measurement performed by medical professionals. Linear regression analysis of the calculated parameters and the measurements manually obtained by the cardiologists resulted in the strongly correlated values: peak systolic velocity's and mean pressure gradient's R2 both equal to 0.99, their means' differences equal to 0.02 m/s and 4.09 mmHg, respectively, and aortic valve area's R2 of 0.89 with the two methods means' difference of 0.19 mm.
Conclusions: The introduced Doppler echocardiography images processing method can be used as a computer-aided assistance in the aortic valve stenosis diagnostics. In our future work, we intend to improve precision of left ventricular outflow tract spectrogram measurements and apply data mining methods to propose a clinical decision support system for diagnosing aortic valve stenosis.
Keywords: Digital image processing; Doppler echocardiography; aortic valve stenosis; feature extraction.