Automatic Landmark Identification in 3D Cone-Beam Computed Tomography scans

Maxime Paul Guy Gillot, Baptiste Antoine Baquero, Juan Carlos Prieto, Lucia Cevidanes

Keywords: Deep learning, Agent based learning, Medical image analysis, Multi-scale images, Three-dimensional landmark identification, Real-time detection, Smart localization.

TL;DR: Automatic Landmark Identification in 3D medical images using a deep neural network agent moving in a multi-scale voxel-grid

Abstract:

Robust and fast solutions for anatomical landmark detection support entire clinical work-flows from diagnosis, therapy planning, intervention and follow-up, image-to-image regis-tration, structure tracking and simulations. In this paper, we propose a novel approach that reformulates landmark detection as a classification problem through a virtual agent placed inside a 3D Cone-Beam Computed Tomography (CBCT) scan.
This agent is trained to navigate in a multi-scale volumetric space to reach the estimated landmark position. The agent movements decision relies on a combination of Densely Connected Convolutional Networks (DCCN) and fully connected layers. We evaluated our approach with 60 CBCTs from teenagers to senior patients.
Each of them have 34 different ground truth landmarks position identified by clinicians.
Our method achieved high accuracy with an average of less than a 1.3mm error on the landmarks position without failures. Moreover, the total computation time to identify 6 landmarks is of 25.2s on large 3D-CBCT scans using GPU.

Registration: I acknowledge that publication of this at MIDL and in the proceedings requires at least one of the authors to register and present the work during the conference.

Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.

Paper Type: both

Primary Subject Area: Detection and Diagnosis

Secondary Subject Area: Application: Radiology

Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.