Matrix decomposition based feature extraction for murmur classification

A Method to Differentiate Pathological Murmurs from Innocent Murmurs in Cardiology

“…The heart sound data was preprocessed to minimize environmental noises and then processed with singular spectrum analysis (SSA), which separated murmurs from recorded heart sounds [2][3][4]. The power spectrum densities (PSD) of extracted murmurs in two groups were estimated.…”

A novel method for discrimination between innocent and pathological heart murmurs

“…Although these studies suggested mathematical methods for the murmur classification, stability of their suggested methods was not validated for the IM class, as IM was not included in their data material. To a lesser extent, IM was considered as the reference group in combination with NM against the pathological murmurs, where the joint wavelet transform and artificial neural network were employed for the screening purpose [6,12,16,40]. However, development of a sophisticated method tailored for IM characterization has often been neglected.…”

“…On the other hand, the non-stationary properties of heart murmurs were mostly ignored in both analysis and validation. An appropriate statistical validation is of particular importance in such methods where dimension of the feature vector is large, that was ignored in the analyses [12,40,41].…”

A hybrid FAM–CART model and its application to medical data classification

“…A number of splitting criteria such as Gini, Symgini, twoing, ordered twoing, and maximum deviance can be used in CART, as explained in [33]. Among them, the Gini index is a key splitting criterion [34] and has been commonly used in undertaking classification problems [33]. It searches for the largest category in a data set and attempts to separate it from other categories, and typically performs better than other criteria [35].…”