Detection of Breast Cancer Using Infrared Thermal Images for Improved Accuracy by Using Random Forest and Multilayer Perceptron

Abstract

At the present time, breast cancer is one of the most often diagnosed forms of cancer in females. Mammography is the most common form of screening imaging used to identify breast cancer in its earlier stages. Nevertheless, thermal infrared pictures (thermography) can be utilized to detect lesions in dense breasts. In this study, the typical areas reflect warmer temperatures than malignant areas. In this study, we offer a unique approach for modeling the temperature variations in normal and abnormal breasts by combining the Random forest and Multilayer perceptron techniques. The project aims to study the accuracy, sensitivity, and specificity of the infrared breast cancer images using infrared thermal images using random forest and multilayer perceptron algorithms and comparing the accuracy, specificity, and sensitivity. Materials and Methods: The information for this study was s gained from thermal images from Visual labs DMR-IR. The samples were considered as (N=60) for Random Forest and (N= 60) for MultiLayer Perceptron. Novel Matlab software is used to calculate accuracy, specificity, and sensitivity. Results: The result demonstrates the accuracy of the thermal breast images using SPSS software. A statistically insignificant difference exists, with Random Forest accuracy (92.5%) with specificity (90%) and with sensitivity (95%) and demonstrated a better outcome in comparison with Multilayer Perceptron accuracy (90%), specificity (91.6%) and sensitivity (88.3%). Conclusion: Random Forest gives better accuracy, specificity, and sensitivity than Multilayer Perceptron to detect breast cancer.