著者

鈴木 裕 深澤 瑞也 森 鷹浩 阪田 治 服部 遊 加藤 隆也

出版者

一般社団法人 電気学会

雑誌

電気学会論文誌C(電子・情報・システム部門誌) (ISSN:03854221)

巻号頁・発行日

vol.130, no.3, pp.401-406, 2010-03-01 (Released:2010-03-01)

参考文献数

20

被引用文献数

4 4

---

It is desired to detect stenosis at an early stage to use hemodailysis shunt for longer time. Stethoscope auscultation of vascular murmurs is useful noninvasive diagnostic approach, but an experienced expert operator is necessary. Some experts often say that the high-pitch murmurs exist if the shunt becomes stenosed, and some studies report that there are some features detected at high frequency by time-frequency analysis. However, some of the murmurs are difficult to detect, and the final judgment is difficult. This study proposes a new diagnosis support system to screen stenosis by using vascular murmurs. The system is performed using artificial neural networks (ANN) with the analyzed frequency data by maximum entropy method (MEM). The author recorded vascular murmurs both before percutaneous transluminal angioplasty (PTA) and after. Examining the MEM spectral characteristics of the high-pitch stenosis murmurs, three features could be classified, which covered 85 percent of stenosis vascular murmurs. The features were learnt by the ANN, and judged. As a result, a percentage of judging the classified stenosis murmurs was 100%, and that of normal was 86%.

著者

脇 隼人 鈴木 裕 阪田 治 深澤 瑞也 加藤 初弘

出版者

一般社団法人 電気学会

雑誌

電気学会論文誌C(電子・情報・システム部門誌) (ISSN:03854221)

巻号頁・発行日

vol.132, no.10, pp.1589-1594, 2012-10-01 (Released:2012-10-01)

参考文献数

23

被引用文献数

1 2

---

The number of dialysis patients is approximately 300,000 and is increasing every year in Japan. A renal failure patient requires a hemodialysis shunt for dialysis to be performed; however, the blood vessels around the hemodialysis shunt may become stenosed. The stethoscope auscultation of vascular murmurs has some use in the assessment of access patency; however, this diagnostic approach is skill dependent. Therefore, a diagnostic support system to detect stenosis is desirable. We developed an auscultating diagnosis support system for assessing hemodialysis shunt stenosis by using a self-organizing map (SOM) and short-time maximum entropy method. In this paper, for the purpose of improving the accuracy of stenosis detection, the Mel-frequency cepstrum coefficient (MFCC)-based hidden Markov model (HMM) was also used. As a result, a high correlation between an SOM system and HMM system was found. Therefore, the credibility of the each system was confirmed. Furthermore, the results indicated that the accuracy of stenosis detection could be improved by combining the SOM and HMM methods.