Regression based myoglobinuria detection from urine image
Md. Imran Khan
Abstract
Now-a-days urine dipstick test is needed to determine the presence of myoglobin in urine and to quantify myoglobin level in urine sophisticated spectrophotometer method is performed. This paper proposes a soft computation based novel regression model that can determine the presence of myoglobin and also can estimate myoglobin level in urine from urine image. Performance analysis of several regression model shows that multivariate nonlinear Hougen regression model provides best fit between estimated myoglobin values with image data. We find that the residuals are near in baseline and the adjusted coefficient of determination (R2) is 0.97 which is very significant. Root mean square error (RMSE) is 19.5 out of 51 urine image samples with 10 error degrees of freedom.
Keyword
Myoglobin, Urine dipstick test, Multivariate nonlinear Hougen regression, Residuals.