Plain-language write-ups of our peer-reviewed work across digital health, AI, and human behaviour.
How a phone became an MHRA-certified Class I medical device that assesses Parkinson's motor symptoms at home — using deep learning to guarantee data quality and cut a 25-minute test to under 4.
Read write-up →A prospective, dual-site study testing whether a 16-item smartphone assessment can predict the motor scores three blinded neurologists give — and what rigorous validation means for digital measures in Parkinson's trials.
Read write-up →An open-source Python toolkit that standardises how Parkinson's symptoms are measured from wearables and smartphones — turning raw sensor data into transparent, comparable biomarkers and clinical scores.
Read write-up →Turning infant brain-activity recordings into images a neural network can read — to predict an autism diagnosis years before it is usually made.
Read write-up →Teaching a neural network to carry knowledge from one financial task to another — using ideas from behavioural genetics and selective breeding to learn faster on new problems.
Read write-up →Two studies of more than 1,200 people measuring how hand strength and precision vary with sex, age, anatomy and lifestyle — and what that reveals about the evolution of the human hand.
Read write-up →A large citizen-science study linking the brain's left/right biases to language ability and social experience — and why a rare 'reversed' profile patterns differently.
Read write-up →A weight-initialisation method for deep neural networks based on interval analysis — aiming to make hard-to-train deep architectures learn more reliably from the start.
Read write-up →