Tuesday, November 28 | 10:00 a.m.-10:10 a.m. | T3-SSMS02-4 | Room N228
CT radiomics analysis can be a viable noninvasive option for aiding management decisions of indeterminate solitary lung nodules detected on noncontrast chest CT scans in colorectal cancer patients, this study has found.
The research team, led by presenter Xinyi Gao, MD, of The Cancer Hospital of the University of Chinese Academy of Sciences, sought to build and validate AI diagnostic models for classifying solitary pulmonary nodules as either pulmonary metastasis, benign lesion, or primary lung cancer. Machine-learning (decision tree, extra trees, light GBM, random forest, support vector machine, and XGBoost) and deep-learning algorithms were used to develop this potential CT radiomics tool.
They trained the model using a random selection of 185 patients with pretreated T1 stage primary lung cancer and 256 patients with benign solitary pulmonary nodules. After testing on an external validation set of 44 colorectal cancer cases from two different hospitals, ,the researchers found that a support vector machine (SVM)-based algorithm and a deep-learning model both offered very high performance.
For classification between pulmonary metastasis and benign solitary pulmonary nodule, the machine-learning model based on a support vector machine algorithm showed the best diagnostic ability with an area under the curve (AUC) of 0.977. Meanwhile, the deep-learning mode was less effective but still had an AUC of 0.893. In classifying between pulmonary metastasis and lung cancer, the machine-learning model had an AUC of 0.991 in internal validation, while the deep-learning model produced an AUC of 0.949.
"Noncontrast CT-based radiomic analyses can be useful for noninvasively differentiating solitary PM and benign pulmonary nodule or primary T1 stage lung cancer which can aid clinical decision in CRC patients with indeterminate solitary pulmonary nodules detected by CT," the authors wrote.
Learn more during the Tuesday morning session.
