Abstract #2911

Physics-Informed Deep Autoencoder for Dynamic Fitting of Deuterium Metabolic Imaging Data

Aaron Paul Osburg¹, Wolfgang Bogner^1,2, Amirmohammad Shamaei³, Bernhard Strasser¹, Fabian Niess¹, Lukas Hingerl¹, Anna Duguid¹, Viola Bader¹, Sabina Naomi Frese¹, Martin Krssak⁴, and Stanislav Motyka^1,2

¹High Field MR Center, Department of Biomedical Imaging and Image-Guided Therapy, Medical University of Vienna, Vienna, Austria, ²Christian Doppler Laboratory for MR Imaging Biomarkers (BIOMAK), Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Austria, ³Department of Electrical and Software Engineering, University of Calgary, Calgary, AB, Canada, ⁴Department of Medicine III, Division of Endocrinology and Metabolism, Medical University of Vienna, Vienna, Austria

Synopsis

Keywords: Deuterium, Deuterium, Deuterium Metabolic Imaging, Spectroscopy

Motivation: Deep learning (DL) based methods have been applied for ultra-fast quantification of proton MRSI data. Combining DL-based methods and dynamic 2D-fitting for analyzing deuterium metabolic imaging (DMI) time-course data can lead to more robust and faster metabolite concentration estimations.

Goal(s): Develop a DL-based pipeline for dynamic fitting of DMI data. Compare the DL approach to LCModel for performance assessment.

Approach: Two different deep autoencoders (DAEs) were trained to fit in-vivo DMI datasets and were compared to LCModel.

Results: Good correlation between DAE fits of (non-)denoised data and LCModel fits of denoised data was found. The DAEs outperformed LCModel fits for non-denoised data.

Impact: DL-based dynamic fitting of DMI data allows ultra-fast quantification of metabolite concentration time-courses in good agreement with LCModel fits of low-rank denoised data. The proposed method is more robust against uncertainties caused by low signal-to-noise ratio (SNR) than LCModel.

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.