Science Advances · 2024 · 28 citations · 72 references
Life SciencesNeurosciencePsychiatryCognitive NeuroscienceComputational NeuroscienceMachine LearningCognitive ImpairmentCognitive ModelingFunctional NeuroimagingPredictive ModelingSymptom PredictionTransfer LearningClinical Validation StudiesCognitive Outcome PredictionCognitive Symptom PredictionMachine Learning In NeuroimagingMachine Learning Prediction ModelsPerformance PredictionTransdiagnostic ProcessesTransfer Learning For Phenotyping
Computational psychiatry seeks to link brain function with symptoms, yet cognitive impairments are transdiagnostic, treatment‑resistant, and linked to poor outcomes, and prior work suggests that thousands of participants may be required for reliable prediction, challenging many patient studies. The study aims to train a transfer‑learning model on UK Biobank functional neuroimaging data to predict cognitive functioning in three transdiagnostic clinical samples. Using a transfer‑learning framework, the authors train a model on UK Biobank functional neuroimaging data and apply it to predict cognition in samples of 101 to 224 participants. The model achieved prediction performance comparable to larger studies, improved classical models by up to 116%, and maintained performance across independent datasets, demonstrating that large population‑level models can enhance cognition prediction in clinical samples.
A primary aim of computational psychiatry is to establish predictive models linking individual differences in brain functioning with symptoms. In particular, cognitive impairments are transdiagnostic, treatment resistant, and associated with poor outcomes. Recent work suggests that thousands of participants may be necessary for the accurate and reliable prediction of cognition, questioning the utility of most patient collection efforts. Here, using a transfer learning framework, we train a model on functional neuroimaging data from the UK Biobank to predict cognitive functioning in three transdiagnostic samples (ns = 101 to 224). We demonstrate prediction performance in all three samples comparable to that reported in larger prediction studies and a boost of up to 116% relative to classical models trained directly in the smaller samples. Critically, the model generalizes across datasets, maintaining performance when trained and tested across independent samples. This work establishes that predictive models derived in large population-level datasets can boost the prediction of cognition across clinical studies.
72
The UK Biobank resource with deep phenotyping and genomic data
Clare Bycroft, Colin Freeman, Desislava Petkova et al. · Nature · 2018 · 9.7K citations · Full text