Abstract
The large amount of training samples required to develop a deep learning brain injury model demands enormous computational resources. Here, we study how a transformer neural network (TNN) of high accuracy can be used to efficiently generate pretraining samples for a convolutional neural network (CNN) brain injury model to reduce computational cost. The samples use synthetic impacts emulating real-world events or augmented impacts generated from limited measured impacts. First, we verify that the TNN remains highly accurate for the two impact types (N = 100 each; \({R}^{2}\) of 0.948–0.967 with root mean squared error, RMSE, ~ 0.01, for voxelized peak strains). The TNN-estimated samples (1000–5000 for each data type) are then used to pretrain a CNN, which is further finetuned using directly simulated training samples (250–5000). An independent measured impact dataset considered of complete capture of impact event is used to assess estimation accuracy (N = 191). We find that pretraining can significantly improve CNN accuracy via transfer learning compared to a baseline CNN without pretraining. It is most effective when the finetuning dataset is relatively small (e.g., 2000–4000 pretraining synthetic or augmented samples improves success rate from 0.72 to 0.81 with 500 finetuning samples). When finetuning samples reach 3000 or more, no obvious improvement occurs from pretraining. These results support using the TNN to rapidly generate pretraining samples to facilitate a more efficient training strategy for future deep learning brain models, by limiting the number of costly direct simulations from an alternative baseline model. This study could contribute to a wider adoption of deep learning brain injury models for large-scale predictive modeling and ultimately, enhancing safety protocols and protective equipment.
Similar content being viewed by others
References
Arrue, P., N. Toosizadeh, H. Babaee, and K. Laksari. Low-rank representation of head impact kinematics: a data-driven emulator. Front. Bioeng. Biotechnol. 8:1–11, 2020.
Bayly, P. V., A. Alshareef, A. K. Knutsen, K. Upadhyay, R. J. Okamoto, A. Carass, J. A. Butman, D. L. Pham, J. L. Prince, K. T. Ramesh, and C. L. Johnson. MR imaging of human brain mechanics in vivo: new measurements to facilitate the development of computational models of brain injury. Ann. Biomed. Eng. 2021. https://doi.org/10.1007/s10439-021-02820-0.
Bian, K., and H. Mao. Mechanisms and variances of rotation-induced brain injury: a parametric investigation between head kinematics and brain strain. Biomech. Model. Mechanobiol. 2020. https://doi.org/10.1007/s10237-020-01341-4.
Carlsen, R. W., A. L. Fawzi, Y. Wan, H. Kesari, and C. Franck. A quantitative relationship between rotational head kinematics and brain tissue strain from a 2-D parametric finite element analysis. Brain Multiphysics. 2:100024, 2021.
Dao, T. T. From deep learning to transfer learning for the prediction of skeletal muscle forces. Med. Biol. Eng. Comput. 57:1049–1058, 2019.
Deck, C., N. Bourdet, A. Trog, F. Meyer, V. Noblet, and R. Willinger. Deep learning method to assess brain injury risk. Int. J. Crashworthiness. 2022. https://doi.org/10.1080/13588265.2022.2130600.
Escarcega, J. D., A. K. Knutsen, R. J. Okamoto, D. L. Pham, and P. V. Bayly. Natural oscillatory modes of 3D deformation of the human brain in vivo. J. Biomech. 119:110259, 2021.
Fahlstedt, M., F. Abayazid, M. B. Panzer, A. Trotta, W. Zhao, M. Ghajari, M. D. Gilchrist, S. Ji, S. Kleiven, X. Li, A. N. Annaidh, and P. Halldin. Ranking and rating bicycle helmet safety performance in oblique impacts using eight different brain injury models. Ann. Biomed. Eng. 49:1097–1109, 2021.
Gabler, L. F., J. R. Crandall, and M. B. Panzer. Assessment of kinematic brain injury metrics for predicting strain responses in diverse automotive impact conditions. Ann. Biomed. Eng. 44:3705–3718, 2016.
Ghazi, K., M. Begonia, S. Rowson, and S. Ji. American Football Helmet Effectiveness Against a Strain-Based Concussion Mechanism. Ann. Biomed. Eng. 50:1498–1509, 2022.
Ghazi, K., S. Wu, W. Zhao, and S. Ji. Instantaneous whole-brain strain estimation in dynamic head impact. J. Neurotrauma. 38:1023–1035, 2021.
Guo, M.-H., Z.-N. Liu, T.-J. Mu, and S.-M. Hu. Beyond self-attention: External attention using two linear layers for visual tasks. 2021.
Hernandez, F., L. C. Wu, M. C. Yip, K. Laksari, A. R. Hoffman, J. R. Lopez, G. A. Grant, S. Kleiven, and D. B. Camarillo. Six degree-of-freedom measurements of human mild traumatic brain injury. Ann. Biomed. Eng. 43:1918–1934, 2015.
Ji, S., M. Ghajari, H. Mao, H. Kraft, Reuben, M. Hajiaghamemar, M. B. Panzer, R. Willinger, M. D. Gilchrist, S. Kleiven, and J. D. Stitzel. Use of brain biomechanical models for monitoring impact exposure in contact sports. Ann. Biomed. Eng. 50:1389–1408, 2022.
Ji, S., S. Wu, and W. Zhao. Dynamic characteristics of impact-induced brain strain in the corpus callosum. Brain Multiphys. 3:100046, 2022.
Ji, S., and W. Zhao. A pre-computed brain response atlas for instantaneous strain estimation in contact sports. Ann. Biomed. Eng. 43:1877–1895, 2015.
Ji, S., and W. Zhao. Displacement voxelization to resolve mesh-image mismatch: application in deriving dense white matter fiber strains. Comput. Methods Programs Biomed. 213:106528, 2022.
Ji, S., W. Zhao, Z. Li, and T. W. McAllister. Head impact accelerations for brain strain-related responses in contact sports: a model-based investigation. Biomech. Model. Mechanobiol. 13:1121–1136, 2014.
King, A. I., K. H. Yang, L. Zhang, W. Hardy, and D. C. Viano. Is head injury caused by linear or angular acceleration? 2003.
Kobeissi, H., S. Mohammadzadeh, and E. Lejeune. Enhancing mechanical metamodels with a generative model-based augmented training dataset. J. Biomech. Eng. 144:121002, 2022.
Kortylewski, A., A. Schneider, T. Gerig, B. Egger, A. Morel-Forster, and T. Vetter. Training Deep Face Recognition Systems with Synthetic Data. 1–8, 2018.
Lin, N., S. Wu, and S. Ji. A morphologically individualized deep learning brain injury model. J. Neurotrauma (in press). 2023. https://doi.org/10.1089/neu.2022.0413.
Liu, Y., A. G. Domel, N. J. Cecchi, E. Rice, A. A. Callan, S. J. Raymond, Z. Zhou, X. Zhan, Y. Li, M. M. Zeineh, G. A. Grant, and D. B. Camarillo. Time window of head impact kinematics measurement for calculation of brain strain and strain rate in American Football. Ann. Biomed. Eng. 2021. https://doi.org/10.1007/s10439-021-02821-z.
Madhukar, A., and M. Ostoja-Starzewski. Finite element methods in human head impact simulations: a review. Ann. Biomed. Eng. 47:1832–1854, 2019.
Meaney, D. F., B. Morrison, and C. R. Bass. The mechanics of traumatic brain injury: a review of what we know and what we need to know for reducing its societal burden. J. Biomech. Eng. 136:021008, 2014.
Menichetti, A., L. Bartsoen, B. Depreitere, J. Vander Sloten, and N. Famaey. A machine learning approach to investigate the uncertainty of tissue-level injury metrics for cerebral contusion. Front. Bioeng. Biotechnol. 9:0201008, 2021.
Miller, L. E., J. E. Urban, M. A. Espeland, M. P. Walkup, J. M. Holcomb, E. M. Davenport, A. K. Powers, C. T. Whitlow, J. A. Maldjian, and J. D. Stitzel. Cumulative strain-based metrics for predicting subconcussive head impact exposure–related imaging changes in a cohort of American youth football players. J. Neurosurg. 29(4):387–396, 2022.
Montanino, A., X. Li, Z. Zhou, M. Zeineh, D. B. Camarillo, and S. Kleiven. Subject-specific multiscale analysis of concussion: from macroscopic loads to molecular-level damage. Brain Multiphys. 2:100027, 2021.
Post, A., E. S. Walsh, T. B. Hoshizaki, and M. D. Gilchrist. Analysis of loading curve characteristics on the production of brain deformation metrics. Proc. Inst. Mech. Eng. Part P J. Sport. Eng. Technol. 0:1–8, 2012.
Sanchez, E. J., L. F. Gabler, A. B. Good, J. R. Funk, J. R. Crandall, and M. B. Panzer. A reanalysis of football impact reconstructions for head kinematics and finite element modeling. Clin. Biomech. 64:82–89, 2018.
Takhounts, E. G., S. A. Ridella, R. E. Tannous, J. Q. Campbell, D. Malone, K. Danelson, J. Stitzel, S. Rowson, and S. Duma. Investigation of traumatic brain injuries using the next generation of simulated injury monitor (SIMon) finite element head model. Stapp Car Crash J. 52:1–31, 2008.
Tremblay, J., A. Prakash, D. Acuna, M. Brophy, V. Jampani, C. Anil, T. To, E. Cameracci, S. Boochoon, and S. Birchfield. Training deep networks with synthetic data: Bridging the reality gap by domain randomization. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit. Work. 2018-June:1082–1090, 2018.
Wu, S., W. Zhao, S. Barbat, J. Ruan, and S. Ji. Instantaneous brain strain estimation for automotive head impacts via deep learning. Stapp Car Crash J. 65:139–162, 2021.
Wu, S., W. Zhao, K. Ghazi, and S. Ji. Convolutional neural network for efficient estimation of regional brain strains. Sci. Rep. 9:17326, 2019.
Wu, S., W. Zhao, and S. Ji. Real-time dynamic simulation for highly accurate spatiotemporal brain deformation from impact. Comput. Methods Appl. Mech. Eng. 394:114913, 2022.
Wu, S., W. Zhao, Z. Wu, J. C. Ford, L. A. Flashman, T. W. McAllister, J. Hu, and S. Ji. Subject-specific Head Injury Models via Scaling Based on Head Morphology: Initial Finding. , 2019.
Yang, K. H., J. Hu, N. A. White, A. I. King, C. C. Chou, and P. Prasad. Development of numerical models for injury biomechanics research: a review of 50 years of publications in the Stapp Car Crash Conference. Stapp Car Crash J. 50:429–490, 2006.
Zhan, X., Y. Li, Y. Liu, N. J. Cecchi, O. Gevaert, M. M. Zeineh, G. A. Grant, and D. B. Camarillo. Piecewise multivariate linearity between kinematic features and cumulative strain damage measure (CSDM) across different types of head impacts. Ann. Biomed. Eng. 2022. https://doi.org/10.1007/s10439-022-03020-0.
Zhan, X., Y. Li, Y. Liu, A. G. Domel, H. V. Alizadeh, S. J. Raymond, J. Ruan, S. Barbat, S. Tiernan, O. Gevaert, and M. M. Zeineh. The relationship between brain injury criteria and brain strain across different types of head impacts can be different. J. R. Soc. Interface. 18(179):20210260, 2021.
Zhan, X., Y. Liu, S. J. Raymond, H. V. Alizadeh, A. G. Domel, O. Gevaert, M. M. Zeineh, G. A. Grant, and D. B. Camarillo. Rapid estimation of entire brain strain using deep learning models. IEEE Trans. Biomed. Eng. 9294:1–11, 2021.
Zhang, C., and S. Ji. Sex differences in axonal dynamic responses under realistic tension using finite element models. J. Neurotrauma (in press), 2023.
Zhao, W., Y. Cai, Z. Li, and S. Ji. Injury prediction and vulnerability assessment using strain and susceptibility measures of the deep white matter. Biomech. Model. Mechanobiol. 16:1709–1727, 2017.
Zhao, W., and S. Ji. Brain strain uncertainty due to shape variation in and simplification of head angular velocity profiles. Biomech. Model. Mechanobiol. 16:449–461, 2017.
Acknowledgements
Funding from the NSF award under Grant No. 2114697 is acknowledged.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare no competing interests.
Additional information
Associate Editor Stefan M. Duma oversaw the review of this article.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Resubmitted to the Annals of Biomedical Engineering (AMBE) August 16, 2023.
Supplementary Information
Below is the link to the electronic supplementary material.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Lin, N., Wu, S., Wu, Z. et al. Efficient Generation of Pretraining Samples for Developing a Deep Learning Brain Injury Model via Transfer Learning. Ann Biomed Eng (2023). https://doi.org/10.1007/s10439-023-03354-3
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10439-023-03354-3