Brain-based ranking of cognitive domains to predict schizophrenia

Abstract

Schizophrenia is a devastating brain disorder that disturbs sensory perception, motor action, and abstract thought. Its clinical phenotype implies dysfunction of various mental domains, which has motivated a series of theories regarding the underlying pathophysiology. Aiming at a predictive benchmark of a catalogue of cognitive functions, we developed a bottom-up machine-learning strategy and provide a proof of principle in a multi-site clinical dataset (n=324). Existing neuroscientific knowledge on diverse cognitive domains was first condensed into neuro-topographical maps. We then examined how the ensuing meta-analytic cognitive priors can distinguish patients and controls using brain morphology and intrinsic functional connectivity. Some affected cognitive domains supported well-studied directions of research on auditory evaluation and social cognition. However, rarely suspected cognitive domains also emerged as disease-relevant, including self-oriented processing of bodily sensations in gustation and pain. Such algorithmic charting of the cognitive landscape can be used to make targeted recommendations for future mental health research.

Application of machine learning methods in predicting schizophrenia and bipolar disorders: A systematic review

“…In this study, 10 of the 13 studies provided details of a validation process for the applied models 24,26,29,30,37–43,46,47 . Different forms of internal cross‐validation and holdout datasets were most commonly applied.…”

“…Accuracy and sensitivity (recall)-specificity were the most frequently reported performance statistics (in 10 references 24,26,29,37,39,40,42,43,46,47 and 7 of studies, 24,26,38,40,42,46,47 respectively). Five of the studies 26,37,40,44,46 reported AUC, while Precision was reported by 2 studies.…”

“…In this study, 10 of the 13 studies provided details of a validation process for the applied models. 24,26,29,30,[37][38][39][40][41][42][43]46,47 Different forms of internal cross-validation and holdout datasets were most commonly applied. K-fold cross-validation and leave-one-out cross-validation frameworks were used to split data into training, validation, testing sets, or optimizing model parameters.…”

From YouTube to the brain: Transfer learning can improve brain-imaging predictions with deep learning

“…ResNets with deeper layers have been shown to be more easily trainable than traditional DNNs with state-of-the-art results on benchmark tests (K. He et al, 2016). Stacking is an ensemble model averaging technique (Hastie et al, 2009) that we have used in our previous research (Karrer et al, 2019). Multiple base models can be used for separate classifications to then blend the results into a top-level, composite model for a final prediction.…”

The Cost of Untracked Diversity in Brain-Imaging Prediction

“…The prediction paradigm holds the promise of improving disease diagnosis, enhancing prognostic estimates, and ultimately paving the way to precision medicine (Brammer, 2009; Bzdok et al, 2021). Machine learning methods are now increasingly adopted for the goal of classifying various conditions, including autism spectrum disorder (ASD) (Heinsfeld et al, 2018; Plitt et al, 2015; Sabuncu et al, 2015; Wolfers et al, 2019), attention-deficit/hyperactivity disorder (ADHD) (Brown et al, 2012; Riaz et al, 2020; Sen et al, 2018; Wang et al, 2018), anxiety (ANX) (Frick et al, 2014; Liu et al, 2015), or schizophrenia (Davatzikos et al, 2005; Karrer et al, 2019; Rozycki et al, 2018; Shen et al, 2010; Yassin et al, 2020). However, there is a large variability in the accuracy reports across studies (Arbabshirani et al, 2017; Pulini et al, 2019; Woo et al, 2017).…”