| Summary: | 博士 === 國立高雄應用科技大學 === 電子工程系 === 106 === Congenital septal defect (CSD) is one type of heart diseases which take a portion in causing death. The presences of CSD can be detected since early birth which requires expensive devices. For a small clinic in a remote area or in the island of developing country, the medical resources are limited. To fill this gap, this dissertation aims to develop a biomedical signal processing algorithm for hole size diagnosis on people with CSD remotely by means of Internet of Things (IoT).
Heart sounds (HS) contains information about diseases which can be extracted using signal processing approaches. Thus, this dissertation is focused on the hole size diagnosis from the HS with CSD cases. Unlike the former approaches which used feature extraction and selection, a novel approach based on the autocorrelation of the second sounds (S2) per segment and cross correlation of the first (S1) and S2 sounds per segment is presented in this dissertation. The main contributions of this dissertation are: (1) a novel automatic hole size diagnosis on CSD called Average Distance Scattered Atom of Eigen Values (ADSAE), (2) a new algorithm for HS data gathering via IoT, which called Cognitive Multi Input Multi Output (CogMIMO). The CogMIMO integrated the Queen Honey Bee Migration (QHBM) and Cooperative MIMO (CMIMO) to improve the network longevity and shorten the delay.
The ADSAE was tested on the two types of CSD. The performances are compared with Support Vector Machine (SVM), Fuzzy Clustering (FC) and Ellipse Model (EM). Based on the obtained results the ADSAE performs better in diagnosing hole size of CSD in terms of accuracy and F1 scores. In addition, the CogMIMO also surpass the Single Input Single Output (SISO), Multi Input Single Output (MISO), and CMIMO in terms of network longevity and delay.
|
|---|
