Radiomics Feature Selection for Lung Cancer Subtyping and Prognosis Prediction: A Comparative Study of Ant Colony Optimization and Simulated Annealing
Year: 2023
Type: article
Abstract: Lung cancer subtyping and prognosis prediction play a critical role in the development of individualised treatment strategies, which is a cornerstone of precision medicine. The field of radiomics, which focuses on the quantitative feature extraction from medical pictures, shows great promise as a means to this end. This paper presents a comparative comparison of two effective optimisation algorithms, Ant Colony Optimization (ACO) and Simulated Annealing (SA), for the goal of radiomics feature selection in lung cancer subtyping and prognosis prediction.The remarkable heterogeneity of lung cancer makes accurate subtyping difficult. Utilising a large number of features extracted from medical imaging, such as CT scans, radiomics is able to detect even the most minute of tumour characteristics. However, because to their abundance, overfitting occurs and model generalizability suffers. Feature selection is crucial to solving this problem.Natural-process-inspired ACO and SA are used to find the best radiomic features to use. Both ACO and SA are heuristic algorithms, however SA takes its cues from the metallurgical annealing process, while ACO is based on the foraging behaviour of ants. Both methods seek to reduce the dimensionality of a problem by identifying a subset of features that yields the best predicted performance.In this study, ACO and SA are applied to a sizable dataset containing information about people with lung cancer, allowing for a thorough comparison of the two methods. Accuracy in subtyping and prognosis prediction are two measures used to assess the outcomes. In addition, feature selection's reliability and durability are evaluated. The results of this study provide important insights for researchers and clinicians who want to improve the accuracy of subtyping and prognosis prediction in the era of personalised medicine by using radiomics feature selection for lung cancer.
Language: English
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: diamond
OpenAlex ID: W7218169890
