Classification of Cancer Tissue With Machine Learning Algorithms Using Microwave Datasets.

Area of Science:

• Bioengineering
• Medical Physics
• Computational Biology

Background:

• Rising cancer incidence necessitates faster diagnostic methods.
• Traditional pathology reports are time-consuming.
• Early cancer detection is critical for treatment success.

Purpose of the Study:

• To develop a rapid method for differentiating cancerous from healthy colon tissue.
• To evaluate microwave measurement techniques for pathological tissue analysis.
• To compare classification algorithms for accuracy in tissue identification.

Main Methods:

• Utilized free-space microwave measurements (18-26 GHz) on colon tissue samples.
• Collected scattering parameters, including reflection and transmission coefficients.
• Trained and tested k-nearest neighbors (KNN), artificial neural networks (ANN), and Support Vector Machines (SVM) algorithms.

Main Results:

• Four distinct datasets were created using various combinations of measurement features.
• The KNN algorithm demonstrated the highest classification accuracy.
• Optimal performance was achieved using reflection, transmission coefficients, and frequency data.

Conclusions:

• Microwave measurements offer a promising, expedited approach for cancer diagnosis.
• The KNN algorithm is effective for classifying colon tissue based on microwave data.
• This technique has the potential to significantly reduce diagnostic turnaround times.