A novel intelligent system based on adjustable classifier models for diagnosing heart sounds

Abstract

Abstract A novel intelligent diagnostic system is proposed to diagnose heart sounds (HSs). The innovations of this system are primarily reflected in the automatic segmentation and extraction of the first complex sound $$({ CS }_{1})$$ ( CS 1 ) and second complex sound $$({ CS }_{2})$$ ( CS 2 ) ; the automatic extraction of the secondary envelope-based diagnostic features $$\gamma _{_1}$$ γ 1 , $$\gamma _{_2}$$ γ 2 , and $$\gamma _{_3}$$ γ 3 from $${ CS }_{1}$$ CS 1 and $${ CS }_{2}$$ CS 2 ; and the adjustable classifier models that correspond to the confidence bounds of the Chi-square ( $$\chi ^{2}$$ χ 2 ) distribution and are adjusted by the given confidence levels (denoted as $$\beta$$ β ). The three stages of the proposed system are summarized as follows. In stage 1, the short time modified Hilbert transform (STMHT)-based curve is used to segment and extract $${ CS }_{1}$$ CS 1 and $${ CS }_{2}$$ CS 2 . In stage 2, the envelopes $${ CS _{1}}_{\mathrm{F_{E}}}$$ C S 1 F E and $${ CS _{2}}_{\mathrm{F_{E}}}$$ C S 2 F E for periods $${ CS }_{1}$$ CS 1 and $${ CS }_{2}$$ CS 2 are obtained via a novel method, and the frequency features are automatically extracted from $${ CS _{1}}_{\mathrm{F_{E}}}$$ C S 1 F E and $${ CS _{2}}_{\mathrm{F_{E}}}$$ C S 2 F E by setting different threshold value ( $$Thv$$ Thv ) lines. Finally, the first three principal components determined based on principal component analysis (PCA) are used as the diagnostic features. In stage 3, a Gaussian mixture model (GMM)-based component objective function $$f_{ et }(\mathbf{x })$$ f et ( x ) is generated. Then, the $$\chi ^{2}$$ χ 2 distribution for component k is determined by calculating the Mahalanobis distance from $${\mathbf{x }}$$ x to the class mean $$\mu _{_k}$$ μ k for component k, and the confidence region of component k is determined by adjusting the optimal confidence level $$\beta _{k}$$ β k and used as the criterion to diagnose HSs. The performance evaluation was validated by sounds from online HS databases and clinical heart databases. The accuracy of the proposed method was compared to the accuracies of other state-of-the-art methods, and the highest classification accuracies of $$99.43\%$$ 99.43 % , $$98.93\%$$ 98.93 % , $$99.13\%$$ 99.13 % , $$99.85\%$$ 99.85 % , $$98.62\%$$ 98.62 % , 99.67 $$\%$$ % and 99.91 $$\%$$ % in the detection of MR, MS, ASD, NM, AS, AR and VSD sounds were achieved by setting $$\beta _{k}(k=1, 2, \ldots , 7)$$ β k ( k = 1 , 2 , … , 7 ) to 0.87,0.65,0.67,0.65,0.67,0.79 and 0.87, respectively.