Performance reserves in brain-imaging-based phenotype prediction.

This study examines the impact of sample size on predicting cognitive and mental health phenotypes from brain imaging via machine learning. Our analysis shows a 3- to 9-fold improvement in prediction performance when sample size increases from 1,000 to 1 M participants. However, despite this increase, the data suggest that prediction accuracy remains worryingly low and far from fully exploiting the predictive potential of brain imaging data. Additionally, we find that integrating multiple imaging modalities boosts prediction accuracy, often equivalent to doubling the sample size. Interestingly, the most informative imaging modality often varied with increasing sample size, emphasizing the need to consider multiple modalities. Despite significant performance reserves for phenotype prediction, achieving substantial improvements may necessitate prohibitively large sample sizes, thus casting doubt on the practical or clinical utility of machine learning in some areas of neuroimaging.

Similar neural pathways link psychological stress and brain-age in health and multiple sclerosis.

Clinical and neuroscientific studies suggest a link between psychological stress and reduced brain health in health and neurological disease but it is unclear whether mediating pathways are similar. Consequently, we applied an arterial-spin-labeling MRI stress task in 42 healthy persons and 56 with multiple sclerosis, and investigated regional neural stress responses, associations between functional connectivity of stress-responsive regions and the brain-age prediction error, a highly sensitive machine learning brain health biomarker, and regional brain-age constituents in both groups. Stress responsivity did not differ between groups. Although elevated brain-age prediction errors indicated worse brain health in patients, anterior insula-occipital cortex (healthy persons: occipital pole; patients: fusiform gyrus) functional connectivity correlated with brain-age prediction errors in both groups. Finally, also gray matter contributed similarly to regional brain-age across groups. These findings might suggest a common stress-brain health pathway whose impact is amplified in multiple sclerosis by disease-specific vulnerability factors.

A cognitive fingerprint in human random number generation.

Is the cognitive process of random number generation implemented via person-specific strategies corresponding to highly individual random generation behaviour? We examined random number sequences of 115 healthy participants and developed a method to quantify the similarity between two number sequences on the basis of Damerau and Levenshtein's edit distance. "Same-author" and "different author" sequence pairs could be distinguished (96.5% AUC) based on 300 pseudo-random digits alone. We show that this phenomenon is driven by individual preference and inhibition of patterns and stays constant over a period of 1 week, forming a cognitive fingerprint.

Promises and pitfalls of deep neural networks in neuroimaging-based psychiatric research.

By promising more accurate diagnostics and individual treatment recommendations, deep neural networks and in particular convolutional neural networks have advanced to a powerful tool in medical imaging. Here, we first give an introduction into methodological key concepts and resulting methodological promises including representation and transfer learning, as well as modelling domain-specific priors. After reviewing recent applications within neuroimaging-based psychiatric research, such as the diagnosis of psychiatric diseases, delineation of disease subtypes, normative modeling, and the development of neuroimaging biomarkers, we discuss current challenges. This includes for example the difficulty of training models on small, heterogeneous and biased data sets, the lack of validity of clinical labels, algorithmic bias, and the influence of confounding variables.

Different scaling of linear models and deep learning in UKBiobank brain images versus machine-learning datasets.

Recently, deep learning has unlocked unprecedented success in various domains, especially using images, text, and speech. However, deep learning is only beneficial if the data have nonlinear relationships and if they are exploitable at available sample sizes. We systematically profiled the performance of deep, kernel, and linear models as a function of sample size on UKBiobank brain images against established machine learning references. On MNIST and Zalando Fashion, prediction accuracy consistently improves when escalating from linear models to shallow-nonlinear models, and further improves with deep-nonlinear models. In contrast, using structural or functional brain scans, simple linear models perform on par with more complex, highly parameterized models in age/sex prediction across increasing sample sizes. In sum, linear models keep improving as the sample size approaches ~10,000 subjects. Yet, nonlinearities for predicting common phenotypes from typical brain scans remain largely inaccessible to the examined kernel and deep learning methods.

Inferring disease subtypes from clusters in explanation space.

Identification of disease subtypes and corresponding biomarkers can substantially improve clinical diagnosis and treatment selection. Discovering these subtypes in noisy, high dimensional biomedical data is often impossible for humans and challenging for machines. We introduce a new approach to facilitate the discovery of disease subtypes: Instead of analyzing the original data, we train a diagnostic classifier (healthy vs. diseased) and extract instance-wise explanations for the classifier's decisions. The distribution of instances in the explanation space of our diagnostic classifier amplifies the different reasons for belonging to the same class-resulting in a representation that is uniquely useful for discovering latent subtypes. We compare our ability to recover subtypes via cluster analysis on model explanations to classical cluster analysis on the original data. In multiple datasets with known ground-truth subclasses, particularly on UK Biobank brain imaging data and transcriptome data from the Cancer Genome Atlas, we show that cluster analysis on model explanations substantially outperforms the classical approach. While we believe clustering in explanation space to be particularly valuable for inferring disease subtypes, the method is more general and applicable to any kind of sub-type identification.

On utilizing uncertainty information in template-based EEG-fMRI ballistocardiogram artifact removal.

The correction of ballistocardiogram artifacts in simultaneous EEG-fMRI often yields unsatisfactory results. To improve the signal-to-noise ratio (SNR) of results, we inferred EEG signal uncertainty from postcorrection artifact residuals and computed the uncertainty-weighted mean of ERPs. Using an uncertainty-weighted mean significantly and consistently reduced both inter- and intrasubject SEM in the analysis of auditory evoked responses (AER, indicated by the N1-P2 complex) and in the effects of an auditory oddball paradigm (N1-P3 complex, standard-deviant difference). SNR increased by 3% on average for the AER amplitude (intrasubject) and 17% on average for the auditory oddball ERP (intersubject). This demonstrates that weighting by uncertainty complements existing artifact correction algorithms to increase SNR in ERPs. More specifically, it is an efficient method to utilize seemingly corrupt (difficult-to-correct) EEG data that might otherwise be discarded.

Analysing humanly generated random number sequences: a pattern-based approach.

In a random number generation task, participants are asked to generate a random sequence of numbers, most typically the digits 1 to 9. Such number sequences are not mathematically random, and both extent and type of bias allow one to characterize the brain's "internal random number generator". We assume that certain patterns and their variations will frequently occur in humanly generated random number sequences. Thus, we introduce a pattern-based analysis of random number sequences. Twenty healthy subjects randomly generated two sequences of 300 numbers each. Sequences were analysed to identify the patterns of numbers predominantly used by the subjects and to calculate the frequency of a specific pattern and its variations within the number sequence. This pattern analysis is based on the Damerau-Levenshtein distance, which counts the number of edit operations that are needed to convert one string into another. We built a model that predicts not only the next item in a humanly generated random number sequence based on the item's immediate history, but also the deployment of patterns in another sequence generated by the same subject. When a history of seven items was computed, the mean correct prediction rate rose up to 27% (with an individual maximum of 46%, chance performance of 11%). Furthermore, we assumed that when predicting one subject's sequence, predictions based on statistical information from the same subject should yield a higher success rate than predictions based on statistical information from a different subject. When provided with two sequences from the same subject and one from a different subject, an algorithm identifies the foreign sequence in up to 88% of the cases. In conclusion, the pattern-based analysis using the Levenshtein-Damarau distance is both able to predict humanly generated random number sequences and to identify person-specific information within a humanly generated random number sequence.