Classification of Cancer Tissue With Machine Learning Algorithms Using Microwave Datasets

The global incidence of cancer-related diseases and mortality continues to rise, making early and accurate diagnosis crucial for effective treatment. A key step in cancer treatment is the identification and removal of tumorous tissue, which is subsequently analyzed through pathological examinations. However, traditional pathology reports can take several weeks or even months to be processed. To expedite diagnosis, this study employs microwave measurement techniques to distinguish between healthy and cancerous colon tissue samples. The free-space measurement method is selected due to its suitability for evaluating sensitive pathological tissues. Measurements are conducted across 201 points within the 18–26 GHz frequency range, capturing the scattering parameters of various tissue types. Based on these measurements, four distinct datasets are created, incorporating features such as reflection coefficients, transmission coefficients, and frequency values. Three widely used classification algorithms—k-nearest neighbors (KNNs), artificial neural networks (ANNs), and Support Vector Machines (SVMs)—are evaluated for their performance on these datasets. The highest classification accuracy is achieved using the KNN algorithm on the dataset containing both reflection and transmission coefficients, along with measurement frequency