| This research is a continuation of earlier work, which assessed the use of thermal infrared (TIR) imaging systems as a potential tool for physicians to diagnose and manage Rheumatoid Arthritis (RA). From that work, TIR images were taken of 18 control subjects and 13 patients diagnosed with RA by a rheumatologist. However, in other to further explore and better improve classification 10 additional control subjects and 13 additional patients were recruited. Temperature measurements were extracted from all the joints in the hand, wrist, elbow, ankle, feet and knee images and statistical tests were done to show which joints were the best indicators of RA and good discriminators between the patients and the control subjects. It was established that the 2nd metacarpophalangeal (MCP), the 3rd metacarpophalangeal (MCP), the ankle and the knee were the joints that showed the greatest statistical difference between control subjects and patients. The dataset series were submitted to the C.5 software to generate a classifier (decision tree). The Min and Max datasets were the best in classifying RA with a sensitivity of 96% and a specificity of 92%. The 2nd and 3rd MCPs proved to be the best joints to classify RA using the Min and Max dataset series confirming the statistical results from earlier work; where these same joints were concluded to be the best indicators of RA. The classifier was used to classify the level of severity of the patient group into LOW, MEDIUM and HIGH classes' successfully. |
