Name
Abstract | ||
|---|---|---|
Transthoracic recordings of Bjork-Shiley Convexo-Concave valve sounds from patients, and the availability of both intact and failing valves from elective explants, are used to develop a non-invasive method for identification of progressive valve failure. In vitro testing of instrumented valves has confirmed a sequence of multiple interactions and impacts of the tilting disc with the inlet and outlet struts at valve closure. These interactions have prompted the development of correlation techniques for grouping of valve sounds. The grouped sets are used in subsequent identification and extraction of features, and classification of valve condition. Valve closing sound data from human transthoracic recordings are used for identifying features in the time-frequency domain. The features are optimized, and used in training algorithms for classification. The training phase establishes the weights or coefficients that, in the testing phase, are applied to the extracted features from each event. The test output is a predicted value of outlet strut condition for each event, or beat, in the test dataset. The predicted values for each beat for a given valve are then used to classify the valve condition. Two classification techniques are presented. The first method is a Volterra expansion of coefficients of the extracted time-frequency domain features. The second method is a neural network approach. The techniques have successfully classified all data sets for which valve condition is known |
Year | DOI | Venue |
|---|---|---|
1994 | 10.1109/CBMS.1994.315997 | Winston-Salem, NC |
Keywords | Field | DocType |
acoustic signal processing,acoustic variables measurement,cardiology,failure (mechanical),learning (artificial intelligence),neural nets,patient monitoring,prosthetics,valves,BSCC heart valves,Bjork-Shiley Convexo-Concave valve sounds,Volterra expansion,acoustic evaluation,algorithm training,correlation techniques,elective explants,failing valves,in vitro testing,inlet struts,instrumented valves,intact valves,multiple interactions,noninvasive method,outlet struts,patients,progressive failure,tilting disc,time-frequency domain,transthoracic recordings,valve closure,valve sounds | Computer vision,Data set,Computer science,Remote patient monitoring,Feature extraction,Artificial intelligence,Beat (music),Time–frequency analysis,Statistical classification,Artificial neural network,Time frequency domain | Conference |
ISSN | Citations | PageRank |
1063-7125 | 0 | 0.34 |
References | Authors | |
0 | 4 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Allen C. Eberhardt | 1 | 0 | 0.34 |
| Charles E. Chassaing | 2 | 0 | 0.34 |
| Rebecca S. Inderbitzen | 3 | 0 | 1.01 |
| D. W. Wieting | 4 | 0 | 0.68 |