Prediction of activation patterns preceding hallucinations in patients with schizophrenia using machine learning with structured sparsity

Abstract

Despite significant progress in the field, the detection of fMRI signal changes during hallucinatory events remains difficult and time-consuming. This article first proposes a machine-learning algorithm to automatically identify resting-state fMRI periods that precede hallucinations versus periods that do not. When applied to whole-brain fMRI data, state-of-the-art classification methods, such as support vector machines (SVM), yield dense solutions that are difficult to interpret. We proposed to extend the existing sparse classification methods by taking the spatial structure of brain images into account with structured sparsity using the total variation penalty. Based on this approach, we obtained reliable classifying performances associated with interpretable predictive patterns, composed of two clearly identifiable clusters in speech-related brain regions. The variation in transition-to-hallucination functional patterns not only from one patient to another but also from one occurrence to the next (e.g., also depending on the sensory modalities involved) appeared to be the major difficulty when developing effective classifiers. Consequently, second, this article aimed to characterize the variability within the prehallucination patterns using an extension of principal component analysis with spatial constraints. The principal components (PCs) and the associated basis patterns shed light on the intrinsic structures of the variability present in the dataset. Such results are promising in the scope of innovative fMRI-guided therapy for drug-resistant hallucinations, such as fMRI-based neurofeedback.

Identifying a neuroanatomical signature of schizophrenia, reproducible across sites and stages, using machine learning with structured sparsity

“…It has been demonstrated that this classifier generates a coherent, parsimonious and interpretable predictive map: a highly desirable characteristic in the scope of predictive signature discovery. For example, Enet‐TV has been successfully used, recently, in the prediction of prehallucination functional MRI patterns in a clinical population of schizophrenia patients .…”

Intrusive Experiences In Post-Traumatic Stress Disorder: Treatment Response Induces Changes In The Effective Connectivity of The Anterior Insula

“…Importantly, because half of the sample did not reach the threshold for a positive response to chemo-facilitated psychotherapy, we expect our findings to also be relevant for future neuromodulation trials for severely impaired PTSD patients. In this vein, a recent study confirmed that such patients could be trained to downregulate amygdalar activity using real time fMRI-based neurofeedback (66).…”

Decoding activity in Broca’s area predicts the occurrence of auditory hallucinations across subjects

“…This appears compatible with prior hypotheses of inner speech as a form of action (35). In contrast, hippocampal or temporoparietal structures, also known to be involved in AVH pathophysiology (6,36–38), possibly by reflecting the spatiotemporal, rich and complex content of these experiences (21,39), were not necessary to reach high decoding performances. Even if anteroposterior dysconnectivity between speech-related areas has been regularly shown to be involved in AVHs (40), this new finding suggests either that (a) the highly variable nature of the information computed by these temporal-hippocampal structures is not stereotyped enough to be decoded using lSVM or (b) that most of the fine-grained relevant information conserved between subjects is encoded in Broca’s BOLD activity.…”