Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Neural networks with personalized training for improved MOLLI T1 mapping

Gkatsoni, Olympia ; Xanthis, Christos G. LU ; Johansson, Sebastian LU ; Heiberg, Einar LU orcid ; Arheden, Håkan LU and Aletras, Anthony H. LU orcid (2025) In BMC Medical Imaging 25(1).
Abstract

Background: The aim of this study was to develop a method for personalized training of Deep Neural Networks by means of an MRI simulator to improve MOLLI native T1 estimates relative to conventional fitting methods. Methods: The proposed Personalized Training Neural Network (PTNN) for T1 mapping was based on a neural network which was trained with simulated MOLLI signals generated for each individual scan, taking into account both the pulse sequence parameters and the heart rate triggers of the specific healthy volunteer. Experimental data from eleven phantoms and ten healthy volunteers were included in the study. Results: In phantom studies, agreement between T1 reference values and those obtained with... (More)

Background: The aim of this study was to develop a method for personalized training of Deep Neural Networks by means of an MRI simulator to improve MOLLI native T1 estimates relative to conventional fitting methods. Methods: The proposed Personalized Training Neural Network (PTNN) for T1 mapping was based on a neural network which was trained with simulated MOLLI signals generated for each individual scan, taking into account both the pulse sequence parameters and the heart rate triggers of the specific healthy volunteer. Experimental data from eleven phantoms and ten healthy volunteers were included in the study. Results: In phantom studies, agreement between T1 reference values and those obtained with the PTNN yielded a statistically significant smaller bias than conventional fitting estimates (-26.69 ± 29.5ms vs. -65.0 ± 33.25ms, p < 0.001). For in vivo studies, T1 estimates derived from the PTNN yielded higher T1 values (1152.4 ± 25.8ms myocardium, 1640.7 ± 30.6ms blood) than conventional fitting (1050.8 ± 24.7ms myocardium, 1597.2 ± 39.9ms blood). For PTNN, shortening the acquisition time by eliminating the pause between inversion pulses yielded higher myocardial T1 values (1162.2 ± 19.7ms with pause vs. 1127.1 ± 19.7ms, p = 0.01 myocardium), (1624.7 ± 33.9ms with pause vs. 1645.4 ± 18.7ms, p = 0.16 blood). For conventional fitting statistically significant differences were found. Conclusions: Compared to T1 maps derived by conventional fitting, PTNN is a post-processing method that yielded T1 maps with higher values and better accuracy in phantoms for a physiological range of T1 and T2 values. In normal volunteers PTNN yielded higher T1 values even with a shorter acquisition scheme of eight heartbeats scan time, without deploying new pulse sequences.

(Less)
Please use this url to cite or link to this publication:
author
; ; ; ; and
organization
publishing date
type
Contribution to journal
publication status
published
subject
keywords
Cardiac MRI, Deep learning, MRI simulator, T mapping
in
BMC Medical Imaging
volume
25
issue
1
article number
245
publisher
BioMed Central (BMC)
external identifiers
  • pmid:40597733
  • scopus:105009713114
ISSN
1471-2342
DOI
10.1186/s12880-025-01769-z
language
English
LU publication?
yes
id
006af7f4-6574-4c8d-923d-6df913229c7c
date added to LUP
2025-10-27 10:00:58
date last changed
2025-12-08 13:54:00
@article{006af7f4-6574-4c8d-923d-6df913229c7c,
  abstract     = {{<p>Background: The aim of this study was to develop a method for personalized training of Deep Neural Networks by means of an MRI simulator to improve MOLLI native T<sub>1</sub> estimates relative to conventional fitting methods. Methods: The proposed Personalized Training Neural Network (PTNN) for T<sub>1</sub> mapping was based on a neural network which was trained with simulated MOLLI signals generated for each individual scan, taking into account both the pulse sequence parameters and the heart rate triggers of the specific healthy volunteer. Experimental data from eleven phantoms and ten healthy volunteers were included in the study. Results: In phantom studies, agreement between T<sub>1</sub> reference values and those obtained with the PTNN yielded a statistically significant smaller bias than conventional fitting estimates (-26.69 ± 29.5ms vs. -65.0 ± 33.25ms, p &lt; 0.001). For in vivo studies, T<sub>1</sub> estimates derived from the PTNN yielded higher T<sub>1</sub> values (1152.4 ± 25.8ms myocardium, 1640.7 ± 30.6ms blood) than conventional fitting (1050.8 ± 24.7ms myocardium, 1597.2 ± 39.9ms blood). For PTNN, shortening the acquisition time by eliminating the pause between inversion pulses yielded higher myocardial T<sub>1</sub> values (1162.2 ± 19.7ms with pause vs. 1127.1 ± 19.7ms, p = 0.01 myocardium), (1624.7 ± 33.9ms with pause vs. 1645.4 ± 18.7ms, p = 0.16 blood). For conventional fitting statistically significant differences were found. Conclusions: Compared to T<sub>1</sub> maps derived by conventional fitting, PTNN is a post-processing method that yielded T<sub>1</sub> maps with higher values and better accuracy in phantoms for a physiological range of T<sub>1</sub> and T<sub>2</sub> values. In normal volunteers PTNN yielded higher T<sub>1</sub> values even with a shorter acquisition scheme of eight heartbeats scan time, without deploying new pulse sequences.</p>}},
  author       = {{Gkatsoni, Olympia and Xanthis, Christos G. and Johansson, Sebastian and Heiberg, Einar and Arheden, Håkan and Aletras, Anthony H.}},
  issn         = {{1471-2342}},
  keywords     = {{Cardiac MRI; Deep learning; MRI simulator; T mapping}},
  language     = {{eng}},
  number       = {{1}},
  publisher    = {{BioMed Central (BMC)}},
  series       = {{BMC Medical Imaging}},
  title        = {{Neural networks with personalized training for improved MOLLI T<sub>1</sub> mapping}},
  url          = {{http://dx.doi.org/10.1186/s12880-025-01769-z}},
  doi          = {{10.1186/s12880-025-01769-z}},
  volume       = {{25}},
  year         = {{2025}},
}