Pre- and post-surgery brain tumor multimodal magnetic resonance imaging data optimized for large scale computational modelling

On the Readiness of Scientific Data Papers for a Fair and Transparent Use in Machine Learning

To ensure the fairness and trustworthiness of machine learning (ML) systems, recent legislative initiatives and relevant research in the ML community have pointed out the need to document the data used to train ML models. Besides, data-sharing practices in many scientific domains have evolved in recent years for reproducibility purposes. In this sense, academic institutions' adoption of these practices has encouraged researchers to publish their data and technical documentation in peer-reviewed publications such as data papers. In this study, we analyze how this broader scientific data documentation meets the needs of the ML community and regulatory bodies for its use in ML technologies. We examine a sample of 4041 data papers of different domains, assessing their coverage and trends in the requested dimensions and comparing them to those from an ML-focused venue (NeurIPS D&B), which publishes papers describing datasets. As a result, we propose a set of recommendation guidelines for data creators and scientific data publishers to increase their data's preparedness for its transparent and fairer use in ML technologies.

Predictors of cognition after glioma surgery: connectotomy, structure-function phenotype, plasticity

Determining preoperatively the maximal extent of resection that would preserve cognitive functions is the core challenge of brain tumour surgery. Over the past decade, the methodological framework to achieve this goal has been thoroughly renewed: the population-level topographically-focused voxel-based lesion-symptom mapping has been progressively overshadowed by machine learning (ML) algorithmics, in which the problem is framed as predicting cognitive outcomes in a patient-specific manner from a typically large set of variables. However, the choice of these predictors is of utmost importance, as they should be both informative and parsimonious. In this perspective, we first introduce the concept of connectotomy: instead of parameterizing resection topography through the status (intact/resected) of a huge number of voxels (or parcels) paving the whole brain in the Cartesian 3D-space, the connectotomy models the resection in the connectivity space, by computing a handful number of networks disconnection indices, measuring how the structural connectivity sustaining each network of interest was hit by the resection. This connectivity-informed reduction of dimensionality is a necessary step for efficiently implementing ML tools, given the relatively small number of patient-examples in available training datasets. We further argue that two other major sources of interindividual variability must be considered to improve the accuracy with which outcomes are predicted: the underlying structure-function phenotype and neuroplasticity, for which we provide an in-depth review and propose new ways of determining relevant predictors. We finally discuss the benefits of our approach for precision surgery of glioma.

High-Order Resting-State Functional Connectivity is Predictive of Working Memory Decline After Brain Tumor Resection

Surgical resection is one of the main treatment options for brain tumors. However, there is a risk of postoperative cognitive deterioration associated with resective surgery. Recent studies suggest that pre-surgery brain dynamics captured using functional Magnetic Resonance Imaging (fMRI) could provide valuable information about the risk of post-surgery cognitive decline. However, most of these studies are based on simple regression analysis of the raw fMRI signals that do not capture the underlying complex brain dynamics. Here, we investigated the role of higher-order functional brain networks in predicting cognitive decline after surgical resection of brain tumors. More specifically, we looked at the predictive power of second-order functional brain networks in estimating post-surgery working memory (WM) performance. Our results show that the second-order functional brain networks can accurately predict the working memory decline in patients with glioma and meningioma tumors. These findings suggest that there is an interesting relationship between pre-surgical higher-order brain dynamics and the risk of cognitive decline after surgery, which could potentially yield a better prognostic marker for treatment planning of brain tumor patients.

Functional and structural reorganization in brain tumors: a machine learning approach using desynchronized functional oscillations

Neuroimaging studies have allowed for non-invasive mapping of brain networks in brain tumors. Although tumor core and edema are easily identifiable using standard MRI acquisitions, imaging studies often neglect signals, structures, and functions within their presence. Therefore, both functional and diffusion signals, as well as their relationship with global patterns of connectivity reorganization, are poorly understood. Here, we explore the functional activity and the structure of white matter fibers considering the contribution of the whole tumor in a surgical context. First, we find intertwined alterations in the frequency domain of local and spatially distributed resting-state functional signals, potentially arising within the tumor. Second, we propose a fiber tracking pipeline capable of using anatomical information while still reconstructing bundles in tumoral and peritumoral tissue. Finally, using machine learning and healthy anatomical information, we predict structural rearrangement after surgery given the preoperative brain network. The generative model also disentangles complex patterns of connectivity reorganization for different types of tumors. Overall, we show the importance of carefully designing studies including MR signals within damaged brain tissues, as they exhibit and relate to non-trivial patterns of both structural and functional (dis-)connections or activity.

Acquisition and Analysis of Excised Neocortex from Pediatric Patients with Focal Cortical Dysplasia Using Mesoscale Diffusion MRI

Non-invasive classification of focal cortical dysplasia (FCD) subtypes remains challenging from a radiology perspective. Quantitative imaging biomarkers (QIBs) have the potential to distinguish subtypes that lack pathognomonic features and might help in defining the extent of abnormal connectivity associated with each FCD subtype. A key motivation of diagnostic imaging is to improve the localization of a "lesion" that can guide the surgical resection of affected tissue, which is thought to cause seizures. Conversely, surgical resections to eliminate or reduce seizures provided unique opportunities to develop magnetic resonance imaging (MRI)-based QIBs by affording long scan times to evaluate multiple contrast mechanisms at the mesoscale (0.5 mm isotropic voxel dimensions). Using ex vivo hybrid diffusion tensor imaging on a 9.4 T MRI scanner, the grey to white matter ratio of scalar indices was lower in the resected middle temporal gyrus (MTG) of two neuropathologically confirmed cases of FCD compared to non-diseased control postmortem fixed temporal lobes. In contrast, fractional anisotropy was increased within FCD and also adjacent white matter tracts. Connectivity (streamlines/mm3) in the MTG was higher in FCD, suggesting that an altered connectivity at the lesion locus can potentially provide a tangible QIB to distinguish and characterize FCD abnormalities. However, as illustrated here, a major challenge for a robust tractographical comparison lies in the considerable differences in the ex vivo processing of bioptic and postmortem samples. Mesoscale diffusion MRI has the potential to better define and characterize epileptic tissues obtained from surgical resection to advance our understanding of disease etiology and treatment.