Joint prediction of multiple scores captures better individual traits from brain images

Abstract

To probe individual variations in brain organization, population imaging relates features of brain images to rich descriptions of the subjects such as genetic information or behavioral and clinical assessments. Capturing common trends across these measurements is important: they jointly characterize the disease status of patient groups. In particular, mapping imaging features to behavioral scores with predictive models opens the way toward more precise diagnosis. Here we propose to jointly predict all the dimensions (behavioral scores) that make up the individual profiles, using so-called multi-output models. This approach often boosts prediction accuracy by capturing latent shared information across scores. We demonstrate the efficiency of multi-output models on two independent resting-state fMRI datasets targeting different brain disorders (Alzheimer's Disease and schizophrenia). Furthermore, the model with joint prediction generalizes much better to a new cohort: a model learned on one study is more accurately transferred to an independent one. Finally, we show how multi-output models can easily be extended to multi-modal settings, combining heterogeneous data sources for a better overall accuracy.

Enhanced estimations of post‐stroke aphasia severity using stacked multimodal predictions

“…Compared to the above literature, STAMP showed higher predicted vs. true correlation ( r = 0.79–0.88) by integrating multimodal information. The advantage of STAMP might derive from the introduction of prediction stacking, an approach which has been recently shown to improve the predictive accuracy over predictions derived from a single model [Rahim et al, ].…”

Joint Prediction of Breast Cancer Histological Grade and Ki-67 Expression Level Based on DCE-MRI and DWI Radiomics

“…In the machine learning community, the joint estimations of Ki-67 expression level and tumor grade in breast cancer can be formulated as a multitask classification problem [36]. Multitask learning methods have been applied in brain images to jointly predict behavioral scores that make up the individual profiles [37], and the results of improved predictions justify the idea. However, whether the prediction accuracy of correlating between combined MRI radiomics and multiple clinical variables is better than that of associating single parametric MR radiomics with one clinical indicator is still an open question.…”

MOU-EC: model-based whole-brain effective connectivity to extract biomarkers for brain dynamics from fMRI data and study distributed cognition

“…Nonetheless, it is expected that a model with poor fit hardly extracts any relevant information. Many recent studies have used connectivity measures to predict which tasks are performed by the subjects in the scanner [75], the pathological conditions of patients [95,119] and individual identity [105,54,26,5]. Machine learning is the standard for identifying biomarkers of neuropathologies [147] and is also widely used in voxel-based analysis for cognitive tasks [109,152].…”

“…This is a problem since the parameters tuning of the model can be influenced by the specific noise of the data samples and, in turn, the results will be not generalizable to new samples -a phenomenon called overfitting. To evaluate the robustness of a classifier, a cross-validation procedure is the standard for voxel-wise studies for activation maps [109,152] and 385 for clinical applications [119,87]. This procedure consists in splitting the data samples in train and test sets that are respectively used to fit the classifier and to assess its performance, as described in Fig.…”