RSNA 2017 Artificial Intelligence Preview

AI exploits tumor imaging features to predict survival

By Erik L. Ridley, staff writer

November 6, 2017 --

Monday, November 27 | 11:50 a.m.-12:00 p.m. | SSC04-09 | Room E353A
Artificial intelligence (AI) can make use of tumor heterogeneity features on MRI to accurately predict the survival of metastatic colon cancer patients, according to a study by Harvard researchers.

Intratumor heterogeneity has been previously shown to be an independent predictor of patient survival in several cancer subtypes, and spatial variations in tumor enhancement on MRI serve as a macroscopic imaging marker of tumor heterogeneity, said Dr. Dania Daye, PhD, from Harvard Medical School in Boston.

The researchers sought to identify the quantitative MRI-based measures of intratumor heterogeneity that can serve as predictors of survival in patients with metastatic colorectal cancer. They also wanted to use machine learning to build a prognostic model to predict outcomes in these patients, Daye said.

In a retrospective study involving 52 patients with stage IV colon cancer who had received MRI scans for liver metastasis evaluation, regression analysis of tumor image features revealed that Tamura texture features, directionality, coarseness, and contrast were all independently associated with patient survival. In addition, a machine-learning model that assessed imaging-based heterogeneity features, as well as clinical and pathological variables from the patient's electronic medical record, was able to predict survival with an area under the curve of 0.98.

"Defining and validating imaging-based heterogeneity features based on the appearance of metastatic lesions has the potential to improve treatment response prediction and assessment for individual lesions," Daye told "This may help guide metastasis-directed treatment strategies and ultimately help further personalize treatment choices in patients with metastatic colon cancer."

Stop by this talk on Monday to get all the details.

Last Updated np 11/2/2017 3:01:58 PM