Statistical modeling of electrode based thermal therapy with Taguchi based multiple regression

• Maximum achieved temperature was used as the quality characteristic to be observed.
• The research takes into account physical as well as tunable predictors.
• Observed values were obtained using Taguchi's L27 orthogonal array.
• Analysis of variance was carried out to quantify the effect of selected predictors.
• Proposed regression model could account for 90% of the variation in data.

In this research the effect of physical predictors for the electrode based thermal therapy were considered and their effect on the final achieved temperature was quantified. For this purpose 6 predictors that may affect the temperature were studied. Knowledge of temperature inside the domain determines the extent of thermal damage in the domain. For the particular case of electrode based thermal therapy, a prediction model based on the regression analysis can be of immense importance in providing a quick and cost effective way to predict approximate temperature that would be achieved inside the computational domain. The regression model based on the physical predictors in an ideal situation can eradicate the need to solve complex electro-thermal equations for the prediction of temperature. In the first phase, a multiple regression model assuming a linear relationship between six independent variables and maximum achieved temperature as dependent variable or response was considered to best fit the observed values. A dummy variable was incorporated in the model to categorize the data for tissue and tumor. It was observed that multiple linear regression model could explain 82% of the variation in the observed data. In order to obtain a better prediction model, regression diagnostics were carried out and a modified model was obtained catering to the inherent nonlinear dependence of response variable on some variables. It was concluded that for electrode based thermal therapy improved regression model provided a marked improvement in terms of reliability of prediction over the original multiple linear regression model. Modified model was able to explain approximately 90% of the variation in observed data compared to 82% variation in data explained by the original model. Finally, it was revealed that as opposed to simulation software, the proposed regression model presents itself as being cheap, economical and quick alternative which can provide an estimate of the maximum temperature achievable inside the biological tissue.