Extensive T1-weighted MRI preprocessing improves generalizability of deep brain age prediction models

被引:4
|
作者
Dular L. [1 ]
Pernuš F. [1 ]
Spiclin [1 ]
机构
[1] University of Ljubljana, Faculty of Electrical Engineering, Tržaška cesta 25, Ljubljana
关键词
Brain age; Dataset bias; Deep regression models; Linear mixed effect models; MRI preprocessing; Reproducible research; Transfer learning; UK Biobank;
D O I
10.1016/j.compbiomed.2024.108320
中图分类号
学科分类号
摘要
Brain age is an estimate of chronological age obtained from T1-weighted magnetic resonance images (T1w MRI), representing a straightforward diagnostic biomarker of brain aging and associated diseases. While the current best accuracy of brain age predictions on T1w MRIs of healthy subjects ranges from two to three years, comparing results across studies is challenging due to differences in the datasets, T1w preprocessing pipelines, and evaluation protocols used. This paper investigates the impact of T1w image preprocessing on the performance of four deep learning brain age models from recent literature. Four preprocessing pipelines, which differed in terms of registration transform, grayscale correction, and software implementation, were evaluated. The results showed that the choice of software or preprocessing steps could significantly affect the prediction error, with a maximum increase of 0.75 years in mean absolute error (MAE) for the same model and dataset. While grayscale correction had no significant impact on MAE, using affine rather than rigid registration to brain atlas statistically significantly improved MAE. Models trained on 3D images with isotropic 1mm3 resolution exhibited less sensitivity to the T1w preprocessing variations compared to 2D models or those trained on downsampled 3D images. Our findings indicate that extensive T1w preprocessing improves MAE, especially when predicting on a new dataset. This runs counter to prevailing research literature, which suggests that models trained on minimally preprocessed T1w scans are better suited for age predictions on MRIs from unseen scanners. We demonstrate that, irrespective of the model or T1w preprocessing used during training, applying some form of offset correction is essential to enable the model's performance to generalize effectively on datasets from unseen sites, regardless of whether they have undergone the same or different T1w preprocessing as the training set. © 2024 The Author(s)
引用
收藏
相关论文
共 50 条
  • [21] Dystonia associated with hyperintense basal ganglia lesions on T1-weighted brain MRI
    Wolz, Martin
    Junghanns, Susann
    Loehle, Matthias
    von Kummer, Ruediger
    Storch, Alexander
    MOVEMENT DISORDERS, 2008, 23 (11) : 1617 - 1619
  • [22] Clinical Evaluation of Silent T1-Weighted MRI and Silent MR Angiography of the Brain
    Holdsworth, Samantha J.
    Macpherson, Sarah J.
    Yeom, Kristen W.
    Wintermark, Max
    Zaharchuk, Greg
    AMERICAN JOURNAL OF ROENTGENOLOGY, 2018, 210 (02) : 404 - 411
  • [23] Automatic brain tissue segmentation method from MRI T1-weighted data
    Miura, Naoki
    Takahashi, Makoto
    Kawashima, Ryuta
    Kitamura, Masaharu
    FUTURE MEDICAL ENGINEERING BASED ON BIONANOTECHNOLOGY, PROCEEDINGS, 2006, : 1065 - +
  • [24] Effects of Image Noise on different MRI Intensity Normalization Methods in T1-weighted Brain MRI
    Wallimann, P.
    Gabry, H.
    Pouymayou, B.
    Mayinger, M.
    Bogowicz, M.
    Guckenberger, M.
    Andratschke, N.
    von Timmeren, J. E.
    Tanadini-Lang, S.
    STRAHLENTHERAPIE UND ONKOLOGIE, 2022, 198 (SUPPL 1) : S174 - S174
  • [25] An age estimation method using brain local features for T1-weighted images
    Graduate School of Information Scienves, Tohoku University, Japan
    不详
    不详
    不详
    Proc. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. EMBS, (666-669):
  • [26] Automatic segmentation of vestibular schwannomas from T1-weighted MRI with a deep neural network
    Hesheng Wang
    Tanxia Qu
    Kenneth Bernstein
    David Barbee
    Douglas Kondziolka
    Radiation Oncology, 18
  • [27] An Age Estimation Method Using Brain Local Features for T1-Weighted Images
    Kondo, Chihiro
    Ito, Koichi
    Wu, Kai
    Sato, Kazunori
    Taki, Yasuyuki
    Fukuda, Hiroshi
    Aoki, Takafumi
    2015 37TH ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY (EMBC), 2015, : 666 - 669
  • [28] Automatic segmentation of vestibular schwannomas from T1-weighted MRI with a deep neural network
    Wang, Hesheng
    Qu, Tanxia
    Bernstein, Kenneth
    Barbee, David
    Kondziolka, Douglas
    RADIATION ONCOLOGY, 2023, 18 (01)
  • [29] Adaptive model initialization and deformation for automatic segmentation of T1-weighted brain MRI data
    Wu, ZJ
    Paulsen, KD
    Sullivan, JM
    IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, 2005, 52 (06) : 1128 - 1131
  • [30] Hyperintense Dentate Nuclei at Precontrast T1-weighted MRI: Gadolinium Deposition or Brain Irradiation?
    Adin, Mehmet Emin
    Yousem, David Mark
    RADIOLOGY, 2018, 288 (02) : 632 - 633