Matthew L. Olson1, Habib Rahbar1, Brenda F. Kurland2, Xiaoyu Chai2, Joshua Usoro1, Constance D. Lehman1, Savannah C. Partridge1
1University of Washington, Seattle, WA, United States; 2Fred Hutchinson Cancer Research Center, Seattle, WA, United States
On diffusion weighted imaging (DWI), malignant lesions display a lower apparent diffusion coefficient (ADC) than benign lesions, making ADC a potentially useful parameter for discriminating benign and malignant lesions. The low spatial resolution and image quality of DWI can limit interobserver reproducibility of ADC measurements. To address this challenge, we developed a semi-automated method for selection of lesion pixels based on DWI thresholding. Because lesions are typically hyperintense on DWI, a threshold can enable discrimination of lesion pixels from normal parenchyma. We compared inter-observer variability of ADC measurements obtained with this semi-automated approach against the standard manual region-of-interest (ROI) method.