Comparison of Spectral and Sparse Feature Extraction Methods for Heart Sounds Classification / Comparación de Métodos de Extracción de Características Espectrales y Dispersas para Clasificación de Sonidos Cardíacos

Abstract Cardiovascular diseases (CVDs) remain the leading cause of morbidity worldwide. The heart sound signal or phonocardiogram (PCG) is the most simple, low-cost, and effective tool to assist physicians in diagnosing CVDs. Advances in signal processing and machine learning have motivated the design of computer-aided systems for heart illness detection based only on the PCG. The objective of this work is to compare the effects of using spectral and sparse features for a classification scheme to detect the presence/absence of a pathological state in a heart sound signal, more specifically, sparse representations using Matching Pursuit with multiscale Gabor time-frequency dictionaries, linear prediction coding, and Mel-frequency cepstral coefficients. This work compares the performance of PCGs classification applying features as a result of averaging the samples or the features for each PCG sound event when feeding a random forest (RF) classifier. For data balancing, random under-sampling and synthetic minority oversampling (SMOTE) methods were applied. Furthermore, we compare the Correlation Feature Selection (CFS) and Information Gain (IG) for the dimensionality reduction. The findings show a SE=93.17 %, SP=84.32 % and ACC=85.9 % when joining MP+LPC+MFCC features set with an AUC=0.969 showing that these features are promising to be used in heart sounds anomaly detection schemes.

Resumen Las enfermedades cardiovasculares (ECVs) han persistido como la principal causa de mortalidad en el mundo. La señal de audio cardiaco o fonocardiograma (FCG) es la herramienta más simple, efectiva y de bajo costo para auxiliar a especialistas diagnosticando ECVs. Los avances en el procesamiento de señales y aprendizaje máquina han motivado el diseño de auscultación y detección computarizada. El objetivo de este trabajo es comparar el uso de características espectrales y dispersas para un sistema de clasificación que detecte la presencia/ausencia de una patología en un audio cardiaco mediante representaciones dispersas usando Matching Pursuit con diccionarios de Gabor tiempo-frequencia, predicción lineal y coeficientes cepstrales Mel. Se crearon 5 conjuntos de características como resultado de combinar las características para cada FCG y se examinó su desempeño usando un clasificador de bosque aleatorio (RF). Se aplicaron métodos de balanceo de muestras basados en sobremuestreo (SMOTE) y submuestreo aleatorio. Se compararon métodos de selección de características por correlación (CFS) y ganancia de información (IG) para reducir la dimensionalidad del conjunto. Los resultados muestran métricas de SE=93.17 %, SP=84.32 % y ACC=85.9 % al juntar los parámetros MP+LPC+MFCC además de una AUC=0.969. El trabajo muestra el potencial de las características espectrales y escasas para la detección de patologías en señales de audio cardiaco.

Fusion of spectrum and image features to identify Glycyrrhizae Radix et Rhizoma from different origins based on hyperspectral imaging technology / 中国中药杂志

To identify Glycyrrhizae Radix et Rhizoma from different geographical origins, spectrum and image features were extracted from visible and near-infrared(VNIR, 435-1 042 nm) and short-wave infrared(SWIR, 898-1 751 nm) ranges based on hyperspectral imaging technology. The spectral features of Glycyrrhizae Radix et Rhizoma samples were extracted from hyperspectral data and denoised by a variety of pre-processing methods. The classification models were established by using Partial Least Squares Discriminate Analysis(PLS-DA), Support Vector Classification(SVC) and Random Forest(RF). Meanwhile, Gray-Level Co-occurrence matrix(GLCM) was employed to extract textural variables. The spectrum and image data were implemented from three dimensions, including VNIR and SWIR fusion, spectrum and image fusion, and comprehensive data fusion. The results indicated that the spectrum in SWIR range performed better classification accuracy than VNIR range. Compared with other four pre-processing methods, the second derivative method based on Savitzky-Golay(SG) smoothing exhibited the best performance, and the classification accuracy of PLS-DA and SVC models were 93.40% and 94.11%, separately. In addition, the PLS-DA model was superior to SVC and RF models in terms of classification accuracy and model generalization capability, which were evaluated by confusion matrix and receiver operating characteristic curve(ROC). Comprehensive data fusion on SPA bands achieved a classification accuracy of 94.82% with only 28 bands. As a result, this approach not only greatly improved the classification efficiency but also maintained its accuracy. The hyperspectral imaging system, a non-invasively, intuitively and quickly identify technology, could effectively distinguish Glycyrrhizae Radix et Rhizoma samples from different origins.

Study on extraction method of Panax notoginseng plots in Wenshan of Yunnan province based on decision tree model / 中国中药杂志

The herbs used as the material for traditional Chinese medicine are always planted in the mountainous area where the natural environment is suitable. As the mountain terrain is complex and the distribution of planting plots is scattered, the traditional survey method is difficult to obtain accurate planting area. It is of great significance to provide decision support for the conservation and utilization of traditional Chinese medicine resources by studying the method of extraction of Chinese herbal medicine planting area based on remote sensing and realizing the dynamic monitoring and reserve estimation of Chinese herbal medicines. In this paper, taking the Panax notoginseng plots in Wenshan prefecture of Yunnan province as an example, the China-made GF-1multispectral remote sensing images with a 16 m×16 m resolution were obtained. Then, the time series that can reflect the difference of spectrum of P. notoginseng shed and the background objects were selected to the maximum extent, and the decision tree model of extraction the of P. notoginseng plots was constructed according to the spectral characteristics of the surface features. The results showed that the remote sensing classification method based on the decision tree model could extract P. notoginseng plots in the study area effectively. The method can provide technical support for extraction of P. notoginseng plots at county level.