A novel support vector machine learning approach using fractal and run-length matrix indicators for identifying nuclear changes in laryngeal cancer

Introduction/Aim: We aimed to propose a novel and innovative concept of a support vector machine learning algorithm that employs fractal and run-length matrix indicators of nuclear structure to identify malignant squamous epithelial cells in laryngeal cancer. Material and Methods: Regions of interest in micrographs of laryngeal cancer and chronic laryngitis were analyzed using the box-counting fractal and run-length matrix textural techniques. For each nucleus, we quantified fractal dimension values, lacunarity, long-run emphasis, and short-run emphasis. These features were used as input data for training and testing the support vector machine model in the "Scikit-learn" library for Python. Results: The support vector machine model produced relatively good performance indicators. The classification accuracy of the model was 0.83, indicating its adequate ability to distinguish cancer cells from non-cancer cells in our sample. The F1 score (the harmonic mean of precision and recall) was 0.83, suggesting a relatively good balance between these two metrics. The value of the Matthews Correlation Coefficient for this model was 0.65, which indicated moderate agreement between the predicted and actual labels and balanced performance across the two classes. Conclusion: The proposed model provides a solid foundation for further developing artificial intelligence systems for signal analysis in cancer research. If the limitations of this concept are addressed, future research can focus on developing a more comprehensive machine-learning model for identifying laryngeal epithelial cancer cells.