Our research

Monitoring brain biomechanics in sports.

Our brain is the most complex system known to us. Road traffic incidents, sporting activities, falls, disease, or surgical interventions can subject our brain to large mechanical loads, which can damage our brain.

At HEAD Lab, we focus on understanding theeffects of mechanical loading on the brainand using this fundamental understanding to develop applied solutions topredict and prevent brain injuries.

Brain exposure monitoring

We develop technologies to estimate mechanical forces applied to the head and brain in sporting and road-traffic incidents, providing objective information for a range of applications that require brain-exposure monitoring.

Sport as a key application

One key application is sporting, where these technologies can be used to measure biomechanical forces that a player has experienced in a match, season, or their career. This information can benefit medics, coaches, parents and guardians, and sport governing bodies.

Instrumented mouthguards are mouthguards equipped with miniature sensors that measure translational and rotational motion of the head during head-acceleration events. We collaborate with sport governing bodies and mouthguard manufacturers to access thousands of such recordings.

We develop fast-running surrogate models for our detailed brain models that predict brain loading in a fraction of a second. Machine learning has been used to build such models. Our aim is to develop novel surrogate models that can address practical problems and work towards their implementation and application in brain-health surveillance systems.

From a hit to a strain map — in seconds

One real impact — a boxing jab — travelling the full pipeline, from mouthguard kinematics to a coloured brain-strain map.

  1. 01

    Data Capture

    Instrumented mouthguards record translational and rotational head kinematics during head acceleration events. Sensors sample at 1 kHz; each impact window spans ~100 ms.

    1 kHz · ~100 ms window
    Loading mouthguard…
    Rotational velocity rad/s
    Linear acceleration g
  2. 02

    Feature extraction

    Per-axis kinematic time series are reduced to four scalar features — range and sqrt(|max|) of resultant rotational velocity and rotational acceleration — and fed to the model.

    Resultant rotational velocity · rad/sResultant rotational acceleration · rad/s²

    ω and α are the resultant rotational velocity and acceleration; each is reduced to two scalars:

    Rotational velocity rangeωmax⁡−ωmin⁡=22.27 rad/s\omega_{\max} - \omega_{\min} = 22.27\,\text{rad/s}
    Root of peak velocity∣ωmax⁡∣=4.72 rad/s\sqrt{\lvert \omega_{\max} \rvert} = 4.72\,\sqrt{\text{rad/s}}
    Rotational acceleration rangeαmax⁡−αmin⁡=5929 rad/s2\alpha_{\max} - \alpha_{\min} = 5929\,\text{rad/s}^2
    Root of peak acceleration∣αmax⁡∣=77.00 rad/s2\sqrt{\lvert \alpha_{\max} \rvert} = 77.00\,\sqrt{\text{rad/s}^2}
  3. 03

    Inference

    A fine-tuned XGBoost regressor returns a max-principal-strain estimate per white-matter region — R2 0.76–0.85, trained on 1 701 elite male rugby impacts. On iOS the models run on-device via Core ML; on the web the archetype impacts are pre-computed with the same Python pipeline, while live-mouthguard mode runs the models in your browser via ONNX Runtime Web.

    • Corpus Callosum – Genu0.00
    • Corpus Callosum – Body0.00
    • Corpus Callosum – Splenium0.00
    • Forceps Minor0.00
    • Forceps Major0.00
    • Cingulum – Cingulate0.00
    • Cingulum – Hippocampus0.00
    • Sup. Longitudinal Fasc.0.00
    • SLF – Temporal0.00
    • Inf. Fronto-Occipital Fasc.0.00
    • Inf. Longitudinal Fasc.0.00
    • Uncinate Fasc.0.00
    • Ant. Thalamic Radiation0.00
    • Corticospinal Tract0.00
    • Brainstem0.00
    • FA Skeleton0.00
    • Whole Brain0.00
  4. 04

    Visualisation

    Each region is rendered as a dense voxel cloud in its anatomical location (JHU white-matter atlas, MNI152 1 mm space) inside a translucent brain shell, and can be isolated individually. Per-region strain is encoded in the same colourmap the iOS app uses — green to yellow through red, with elevated regions in violet.

    Open the full demo
    Loading anatomy…

Citing trace*

The model behind the demo is described in full in:

Chan, E. Y. K., Yu, X., Qin, C. & Ghajari, M. (2025).Balancing efficiency and accuracy: extreme gradient boosting and neural networks for near real-time brain deformation prediction in sports collisions.Engineering Applications of Artificial Intelligence 162, 112489.doi:10.1016/j.engappai.2025.112489(open access, CC BY-NC-ND 4.0).

Funding

This work has been funded by Sports and Wellbeing Analytics, Cellbond Impact Solutions, the Royal Academy of Engineering Senior Research Fellowship, and the MRC TBI-REPORTER programme.

Try the demo