All Stanford University & Hospital ID holders are now welcome to visit Lane Library! Learn More
Today's Hours: 8:00am - 8:00pm
Filters applied
Did You Mean?
  • Book
    Jiajing Xu.
    A crucial challenge for radiology is maintaining high interpretation accuracy in the face of increasing imaging workload and limited time to review and interpret the images for each patient. Variation in interpretation accuracy among radiologists is a recognized challenge. Two techniques that could help radiologists improve their interpretation are Content-based Image Retrieval (CBIR) for diagnostic support, and lesion size tracking for evaluating response to treatment. CBIR provides assists radiologists to find images from a database that are similar in terms of shared imaging features to the images they are interpreting. The performance of CBIR hinges critically on features that characterize lesions. In the first half of the talk, I will focus on the development of a novel quantitative imaging feature that describes the margin characteristics of lesions. This new margin sharpness feature is robust to variation in lesion segmentation, and achieves excellent CBIR performance in clinical datasets. Tracking lesion size in response to treatment is a crucial component for patient management as well as towards finding the best cancer therapy through clinical trials. Traditionally, tracking lesion size using serial Computed Tomography (CT) scans is largely a manual and tedious process. In the second half of the talk, I will present a novel method to automatically track and segment lymph nodes in serial CT scans. My method has achieved excellent overall segmentation performance compared to manual segmentation provided by radiologists. Ultimately, I envision that the translation of both of the above methods to the clinic will improve diagnostic accuracy, precision, and efficiency.
    Digital Access   2013