Create Combined Thermometric Datasets for Machine Learning in Medicine

Methods for processing medical measurements face a big problem of quality of initial data. Complicating factors are the small size of such samples and the data heterogeneity due to the difficulty o f controlling medical measurements. A promising approach may be the creation of combined dataset, when real measurements are supplemented by the results of simulation modeling of the measurement procedure. Such computer models should reproduce the main physical processes inside biological tissues or organs. As a result, we can basically control the quality of simulation data. The size of the combined dataset can be quite large if there is a good fit between the medical s ample a nd the results of the numerical simulations. This approach is used by us in the diagnosis of breast cancer based on the Microwave Radiometry method. We use various machine learning algorithms to classify “Sick vs. Healthy” on the combined dataset. The analysis shows that the expansion of data due to the modeling results increases the efficiency of diagnosing breast cancer.