The use of smart textiles within clothing offers the facility to monitor patient vital signs in an unobtrusive manner. In the present study we examine the benefits of integrating electrodes into smart shirts taking into consideration aspects of practical limitations in sensor placement. Three practical scenarios are investigated which restrict possible recording sites to the anterior, lateral, and posterior regions, respectively. A wrapper approach incorporating both nearest neighbor and logistic regression models was adopted to search for and extract relevant features. Two discrimination tasks were investigated; identifying between subjects with evidence of old myocardial infarction, and normal healthy subjects; and identifying between subject suffering from left ventricular hypertrophy and healthy subjects. The results from the study indicate that acceptable classification performance is possible even if recording sites are restricted due to practical constraints.
|Journal||International Journal of Assistive Robotics and Mechatronics|
|Publication status||Published - 1 Jun 2007|