| Mostafa Zahed, Janet Kireta, Victor Aderanti, Maryam Skafyan |
Cancer remains a critical global health issue, with early detection and diagnosis essential for successful treatment. This research explores the integration of Radiomics techniques, utilizing advanced imaging modalities such as CT and PET-CT, to enhance the precision of medical image analysis for cancer detection and staging. Radiomics involves extracting quantitative features from medical images and applying machine learning for disease prediction, addressing the challenge of high-dimensional data through effective dimension reduction techniques. Utilizing datasets from medical universities in China, including a 2022 dataset from Harbin Medical University with CT and PET-CT images of 131 lung cancer patients, this study demonstrates the application of dimension reduction methods like PCA, K-means clustering, and Hierarchical Clustering using R, SAS, and Python. These techniques successfully reduced the number of features from 110 to 39 with PCA and to 21 with clustering. Texture-based features, particularly those in the second-order radiomics category, were identified as the most significant for lung cancer diagnosis and prediction. This research underscores the potential of Radiomics and unsupervised learning techniques to optimize feature extraction and reduction processes, enhancing the interpretability and efficiency of cancer prediction models from medical images. The findings highlight the crucial role of texture-based features in improving cancer diagnosis and treatment planning, providing valuable insights for future research and clinical applications. |