197 - Cancer outcomes of head and neck cancer predicted by multi approaches using machine learning approaches

• Published in: Journal of medical imaging and radiation sciences Vol. 55; no. 3
• Main Authors: Chan, Ms Chiu Yan, Li, Ms Tsoi Yin, Lee, Ms Wai Yee, Tang, Dr Fuk Hay
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01-10-2024
• ISSN: 1939-8654
• DOI: 10.1016/j.jmir.2024.101509

Head and neck cancer (HNC) has been causing several thousands of deaths annually in Hong Kong. Accurate cancer prognosis can allow physicians to design a personalized treatment to lower the mortality rate. Significant breakthroughs in Artificial Intelligence (AI) technology have opened a lot of opportunities in the development of cancer prognosis. Traditionally, it has to be done with the consideration of TNM staging and biopsy is required. With the advancement in computer technology, invasive procedures can be skipped with the application of radiomics. Recent research has successfully demonstrated that the use of radiomics and machine learning could indeed achieve an accurate result in cancer prognosis prediction.
In our research, we adopt multiple machine learning approaches, that include Random Forests, deep learning technique, Gradient Boosting and clinical factors for evaluation of cancer outcomes. The clinical data were download from TCIA image database. We use 3D slicer and program developed by the research team to extract multi factors.
The performance of these machine learning approaches can vary depending on the quality and size of the dataset, the choice of features, and the specific problem being addressed. The proposed model will be evaluated under the receiver operating characteristic (ROC) curve to understand its sensitivity, specificity and accuracy.
The Multi-approaches method offer a promising avenue for predicting cancer outcomes in head and neck cancer, and ongoing research in this field continues to improve the accuracy and reliability of these predictions.