Sensor Data Classification for Renal Dysfunction Patients Using Support Vector Machine

Abstract

Breath analysis is a non-invasive technique used in clinical laboratories for the identification of several morbid conditions. Renal failure patients can be distinguished by observing the degree of ammonia in exhaled breath. Standard treatment for renal failure patients is haemodialysis. Ammonia as a biomarker can also provide the efficiency of dialysis. This study proposes a system that uses metal oxide semiconductor sensors, which is capable of detecting ammonia from exhaled breath and identify whether the patient has been subjected to renal failure. Breath samples are gathered using a specially designed breath collecting tube and the steady-state responses are recorded after the samples have been passed through sensors TGS2444, MQ135, MQ137, and TGS826. From the steady-state response, geometric features are extracted and the support vector machine technique is used for classifying the pre- and post-dialysis groups. The results shown an accuracy of 88 % for the designed classifier for sensor TGS2444.