Meeting Banner
Abstract #1553

CNN-Certainty-Directed Utilization of Deep Learning in Radiological Assessment of Adnexal Masses on Pelvic MRI

David Bonekamp1,2,3,4, Theresa Mokry5, Nils Netzer1,3, Thomas Hielscher1, Christa Flechtenmacher5, Lisa Katharina Nees5, Oliver Zivanovic5, Hans-Ulrich Kauczor5, and Heinz-Peter Schlemmer1,2,4
1DKFZ, Heidelberg, Germany, 2National Center for Tumor Diseases (NCT), Heidelberg, Germany, 3Heidelberg University Medical School, Heidelberg, Germany, 4German Cancer Consortium (DKTK), Heidelberg, Germany, 5University Hospital Heidelberg, Heidelberg, Germany

Synopsis

Keywords: Diagnosis/Prediction, Reproductive

Motivation: Multi-parametric pelvic MRI for evaluation of adnexal masses is currently assessed using the O-RADS system. AI-driven approaches using convolutional neural networks (CNN) require further evaluation.

Goal(s): Determine if methods to infer the uncertainty of CNN assessment provide guidance for utilization of fully-automated CNN-based assessment of pelvic MRI to partily substitute for experienced radiologist assessment and for enhanced efficiency of the diagnostic pathway.

Approach: Construct a CNN ensemble for the segmentation and classification of adnexal masses to stratify CNN predictions by CNN prediction certaintly.

Results: A proportion of cases can be assessed by CNN only with no or only minimal reduction in diagnostic accuracy.

Impact: CNN-based triage of multi-parametric pelvic MRI for assessment of adnexal masses has potential to support radiological decision making.

How to access this content:

For one year after publication, abstracts and videos are only open to registrants of this annual meeting. Registrants should use their existing login information. Non-registrant access can be purchased via the ISMRM E-Library.

After one year, current ISMRM & ISMRT members get free access to both the abstracts and videos. Non-members and non-registrants must purchase access via the ISMRM E-Library.

After two years, the meeting proceedings (abstracts) are opened to the public and require no login information. Videos remain behind password for access by members, registrants and E-Library customers.

Click here for more information on becoming a member.

Keywords