Brain age revisited: Investigating the state vs. trait hypotheses of EEG-derived brain-age dynamics with deep learning DOI Creative Commons
Lukas Gemein, Robin Tibor Schirrmeister, Joschka Boedecker

et al.

Imaging Neuroscience, Journal Year: 2024, Volume and Issue: 2, P. 1 - 22

Published: Jan. 1, 2024

Abstract The brain’s biological age has been considered as a promising candidate for neurologically significant biomarker. However, recent results based on longitudinal magnetic resonance imaging (MRI) data have raised questions its interpretation. A central question is whether an increased of the brain indicative pathology and if changes in correlate with diagnosed (state hypothesis). Alternatively, could discrepancy be stable characteristic unique to each individual (trait hypothesis)? To address this question, we present comprehensive study aging clinical Electroencephalography (EEG), which complementary previous MRI-based investigations. We apply state-of-the-art temporal convolutional network (TCN) task regression. train recordings Temple University Hospital EEG Corpus (TUEG) explicitly labeled non-pathological evaluate subjects well pathological recordings, both examinations at single point time TUH Abnormal (TUAB) repeated over time. Therefore, created four novel subsets TUEG that include multiple recordings: (RNP): all non-pathological; (RP): pathological; transition non-patholoigical (TNPP): least one recording followed by (TPNP): similar TNPP but opposing (first then non-pathological). show our TCN reaches performance decoding TUAB mean absolute error 6.6 years R2 score 0.73. Our extensive analyses demonstrate model underestimates subjects, latter significantly (-1 -5 years, paired t-test, p = 0.18 6.6e−3). Furthermore, there exist differences average gap between (RNP vs. RP) (-4 -7.48 permutation test, 1.63e−2 1e−5). find mixed regarding significance classification While it datasets RNP versus RP (61.12% 60.80% BACC, 1.32e−3 1e−5), not TPNP (44.74% 47.79% 0.086 0.483). Additionally, these scores are clearly inferior ones obtained from direct 86% BACC higher. evidence change status within relates (0.46 1.35 0.825 0.43; Wilcoxon-Mann-Whitney Brunner-Munzel 0.13). findings, thus, support trait rather than state hypothesis estimates derived EEG. In summary, findings indicate neural underpinnings likely more multifaceted previously thought, taking into account will benefit interpretation empirically observed dynamics.

Language: Английский

Predicting age from resting-state scalp EEG signals with deep convolutional neural networks on TD-brain dataset DOI Creative Commons

Mariam Khayretdinova,

Alexey Shovkun,

V. N. Degtyarev

et al.

Frontiers in Aging Neuroscience, Journal Year: 2022, Volume and Issue: 14

Published: Dec. 6, 2022

Introduction Brain age prediction has been shown to be clinically relevant, with errors in its associated various psychiatric and neurological conditions. While the from structural functional magnetic resonance imaging data feasible high accuracy, whether same results can achieved electroencephalography is unclear. Methods The current study aimed create a new deep learning solution for brain using raw resting-state scalp EEG. To this end, we utilized TD-BRAIN dataset, including 1,274 subjects (both healthy controls individuals disorders, total of 1,335 recording sessions). achieve best prediction, used augmentation techniques increase diversity training set developed convolutional neural network model. Results model’s was done 10-fold cross-subject cross-validation, EEG recordings not considered test In training, relative rather than absolute loss function led better mean error 5.96 years cross-validation. We found that performance could when both eyes-open eyes-closed states are simultaneously. frontocentral electrodes played most important role prediction. Discussion architecture method proposed networks (DCNN) improve state-of-the-art metrics task by 13%. Given might potential biomarker numerous diseases, inexpensive precise EEG-based estimation will demand clinical practice.

Language: Английский

Citations

20
Prediction of brain age using structural magnetic resonance imaging: A comparison of accuracy and test-retest reliability of publicly available software packages DOI Creative Commons
Ruben P. Dörfel,

Joan M. Arenas‐Gomez,

Patrick M. Fisher

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Jan. 27, 2023

Abstract Background Brain age prediction algorithms using structural magnetic resonance imaging (MRI) aim to assess the biological of human brain. The difference between a person’s chronological and estimated brain is thought reflect deviations from normal aging trajectory, indicating slower, or accelerated, process. Several pre-trained software packages for predicting are publicly available. In this study, we perform head-to-head comparison such with respect 1) predictive accuracy, 2) test-retest reliability, 3) ability track progression over time. Methods We evaluated six packages: brainageR, DeepBrainNet, brainage, ENIGMA, pyment, mccqrnn. accuracy reliability were assessed on MRI data 372 healthy people aged 18.4 86.2 years (mean 38.7 ± 17.5 years). Results All showed significant correlations predicted (r = 0.66 0.97, p < 0.001), pyment displaying strongest correlation. mean absolute error was 3.56 (pyment) 9.54 (ENIGMA). mccqrnn superior in terms (ICC values 0.94 - 0.98), as well longer time span. Conclusion Of packages, brainageR consistently highest reliability.

Language: Английский

Citations

13
Bio-psycho-social factors’ associations with brain age: a large-scale UK Biobank diffusion study of 35,749 participants DOI Creative Commons
Max Korbmacher, Tiril P. Gurholt, Ann‐Marie G. de Lange

et al.

Frontiers in Psychology, Journal Year: 2023, Volume and Issue: 14

Published: June 9, 2023

Brain age refers to predicted by brain features. has previously been associated with various health and disease outcomes suggested as a potential biomarker of general health. Few previous studies have systematically assessed variability derived from single multi-shell diffusion magnetic resonance imaging data. Here, we present multivariate models approaches how they relate bio-psycho-social variables within the domains sociodemographic, cognitive, life-satisfaction, well lifestyle factors in midlife old (

Language: Английский

Citations

13
Examining the reliability of brain age algorithms under varying degrees of participant motion DOI Creative Commons
Jamie L. Hanson,

Dorthea J Adkins,

Eva Bacas

et al.

Brain Informatics, Journal Year: 2024, Volume and Issue: 11(1)

Published: April 4, 2024

Abstract Brain age algorithms using data science and machine learning techniques show promise as biomarkers for neurodegenerative disorders aging. However, head motion during MRI scanning may compromise image quality influence brain estimates. We examined the effects of on predictions in adult participants with low, high, no scans ( Original N = 148; Analytic 138 ). Five popular were tested: brainageR, DeepBrainNet, XGBoost, ENIGMA, pyment. Evaluation metrics, intraclass correlations (ICCs), Bland–Altman analyses assessed reliability across conditions. Linear mixed models quantified effects. Results demonstrated significantly impacted estimates some algorithms, ICCs dropping low 0.609 errors increasing up to 11.5 years high scans. DeepBrainNet pyment showed greatest robustness (ICCs 0.956–0.965). XGBoost brainageR had largest (up 13.5 RMSE) bias motion. Findings indicate artifacts significant ways. Furthermore, our results suggest certain like be preferable deployment populations where acquisition is likely. Further optimization validation critical use a biomarker relevant clinical outcomes.

Language: Английский

Citations

4
Brain age revisited: Investigating the state vs. trait hypotheses of EEG-derived brain-age dynamics with deep learning DOI Creative Commons
Lukas Gemein, Robin Tibor Schirrmeister, Joschka Boedecker

et al.

Imaging Neuroscience, Journal Year: 2024, Volume and Issue: 2, P. 1 - 22

Published: Jan. 1, 2024

Abstract The brain’s biological age has been considered as a promising candidate for neurologically significant biomarker. However, recent results based on longitudinal magnetic resonance imaging (MRI) data have raised questions its interpretation. A central question is whether an increased of the brain indicative pathology and if changes in correlate with diagnosed (state hypothesis). Alternatively, could discrepancy be stable characteristic unique to each individual (trait hypothesis)? To address this question, we present comprehensive study aging clinical Electroencephalography (EEG), which complementary previous MRI-based investigations. We apply state-of-the-art temporal convolutional network (TCN) task regression. train recordings Temple University Hospital EEG Corpus (TUEG) explicitly labeled non-pathological evaluate subjects well pathological recordings, both examinations at single point time TUH Abnormal (TUAB) repeated over time. Therefore, created four novel subsets TUEG that include multiple recordings: (RNP): all non-pathological; (RP): pathological; transition non-patholoigical (TNPP): least one recording followed by (TPNP): similar TNPP but opposing (first then non-pathological). show our TCN reaches performance decoding TUAB mean absolute error 6.6 years R2 score 0.73. Our extensive analyses demonstrate model underestimates subjects, latter significantly (-1 -5 years, paired t-test, p = 0.18 6.6e−3). Furthermore, there exist differences average gap between (RNP vs. RP) (-4 -7.48 permutation test, 1.63e−2 1e−5). find mixed regarding significance classification While it datasets RNP versus RP (61.12% 60.80% BACC, 1.32e−3 1e−5), not TPNP (44.74% 47.79% 0.086 0.483). Additionally, these scores are clearly inferior ones obtained from direct 86% BACC higher. evidence change status within relates (0.46 1.35 0.825 0.43; Wilcoxon-Mann-Whitney Brunner-Munzel 0.13). findings, thus, support trait rather than state hypothesis estimates derived EEG. In summary, findings indicate neural underpinnings likely more multifaceted previously thought, taking into account will benefit interpretation empirically observed dynamics.

Language: Английский

Citations

4