NNFit: A Self-Supervised Deep Learning Method for Accelerated Quantification of High- Resolution Short Echo Time MR Spectroscopy Datasets.

in Radiology. Artificial intelligence by Alexander S Giuffrida, Sulaiman Sheriff, Vicki Huang, Brent D Weinberg, Lee A D Cooper, Yuan Liu, Brian J Soher, Michael Treadway, Andrew A Maudsley, Hyunsuk Shim

TLDR

  • The study develops and evaluates a self-supervised deep-learning method for quantifying EPSI datasets, which offers comparable performance to conventional methods with faster processing times.
  • The method, called NNFit, uses a deep neural network to generate metabolite maps and outperforms a widely used parametric-modeling spectral quantitation method.

Abstract

Purpose To develop and evaluate the performance of NNFit, a self-supervised deep-learning method for quantification of high-resolution short echo-time (TE) echo-planar spectroscopic imaging (EPSI) datasets, with the goal of addressing the computational bottleneck of conventional spectral quantification methods in the clinical workflow. Materials and Methods This retrospective study included 89 short-TE whole-brain EPSI/GRAPPA scans from clinical trials for glioblastoma (Trial 1, May 2014-October 2018) and major-depressive-disorder (Trial 2, 2022- 2023). The training dataset included 685k spectra from 20 participants (60 scans) in Trial 1. The testing-dataset included 115k spectra from 5 participants (13 scans) in Trial 1 and 145k spectra from 7 participants (16 scans) in Trial 2. A comparative analysis was performed between NNFit and a widely used parametric-modeling spectral quantitation method (FITT). Metabolite maps generated by each method were compared using the structural- similarity-index-measure (SSIM) and linear-correlation-coefficient (R). Radiation treatment volumes for glioblastoma based on the metabolite maps were compared with the Dice-coefficient and a two-tailedtest. Results Average SSIM andscores for Trial 1 test set data were 0.91/0.90 (choline), 0.93/0.93 (creatine), 0.93/0.93 (-acetylaspartate), 0.80/0.72 (myo-inositol), and 0.59/0.47 (glutamate + glutamine). Average scores for Trial 2 test set data were 0.95/0.95, 0.98/0.97, 0.98/0.98, 0.92/0.92, and 0.79/0.81 respectively. The treatment volumes had average Dice coefficient of 0.92. NNFit's average processing time was 90.1 seconds, whereas FITT took 52.9 minutes on average. Conclusion This study demonstrates that a deep learning approach to spectral quantitation offers comparable performance to conventional quantification methods for EPSI data, but with faster processing at short-TE. ©RSNA, 2025.

Overview

  • The purpose of this study is to develop and evaluate a self-supervised deep-learning method called NNFit for quantifying high-resolution short echo-time (TE) echo-planar spectroscopic imaging (EPSI) datasets.
  • The study used 89 short-TE whole-brain EPSI/GRAPPA scans from two clinical trials: glioblastoma (Trial 1) and major depressive disorder (Trial 2).
  • The primary objective of the study is to address the computational bottleneck of conventional spectral quantification methods in the clinical workflow.

Comparative Analysis & Findings

  • A comparative analysis was performed between NNFit and a widely used parametric-modeling spectral quantitation method (FITT) to evaluate metabolite maps generated by each method.
  • The study used two metrics to compare the metabolite maps: structural similarity index measure (SSIM) and linear correlation coefficient (R).
  • The results showed that NNFit had comparable performance to FITT for both Trials 1 and 2, with average SSIM and R scores ranging from 0.59 to 0.98.

Implications and Future Directions

  • The study demonstrates the potential of deep learning for spectral quantitation in EPSI data, offering comparable performance to conventional methods with faster processing times.
  • The results suggest that NNFit could be used to improve clinical workflows and reduce processing times for EPSI data.
  • Future studies could explore the use of NNFit in different clinical applications and the development of more advanced deep learning models for spectral quantitation.
Read Full Article