We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

eess.IV

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Electrical Engineering and Systems Science > Image and Video Processing

Title: Single volume lung biomechanics from chest computed tomography using a mode preserving generative adversarial network

Abstract: Local tissue expansion of the lungs is typically derived by registering computed tomography (CT) scans acquired at multiple lung volumes. However, acquiring multiple scans incurs increased radiation dose, time, and cost, and may not be possible in many cases, thus restricting the applicability of registration-based biomechanics. We propose a generative adversarial learning approach for estimating local tissue expansion directly from a single CT scan. The proposed framework was trained and evaluated on 2500 subjects from the SPIROMICS cohort. Once trained, the framework can be used as a registration-free method for predicting local tissue expansion. We evaluated model performance across varying degrees of disease severity and compared its performance with two image-to-image translation frameworks - UNet and Pix2Pix. Our model achieved an overall PSNR of 18.95 decibels, SSIM of 0.840, and Spearman's correlation of 0.61 at a high spatial resolution of 1 mm3.
Comments: 5 pages, 5 figures
Subjects: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
Cite as: arXiv:2110.07878 [eess.IV]
  (or arXiv:2110.07878v1 [eess.IV] for this version)

Submission history

From: Muhammad Faizyab Ali Chaudhary [view email]
[v1] Fri, 15 Oct 2021 06:17:52 GMT (2001kb,D)

Link back to: arXiv, form interface, contact.