Biology subjects

Toschi, N.

Publications and source records attributed to Toschi, N..

7 recordsLinked to original sources

bioRxiv2024

Generation of synthetic TSPO PET maps from structural MRI images

BackgroundNeuroinflammation, a pathophysiological process involved in numerous disorders, is typically imaged using [11C]PBR28 (or TSPO) PET. However, this technique is limited by high costs and ionizing radiation, restricting its widespread clinical use. MRI, a more accessible alternative, is commonly used for structural or functional imaging, but when used using traditional approaches has limited sensitivity to specific molecular processes. This study aims to develop a deep learning model to generate TSPO PET images from structural MRI data collected in human subjects. MethodsA total of 204 scans, from participants with knee osteoarthritis (n = 15 scanned once, 15 scanned twice, 14 scanned three times), back pain (n = 40 scanned twice, 3 scanned three times), and healthy controls (n=28, scanned once), underwent simultaneous 3T MRI and [11C]PBR28 TSPO PET scans. A 3D U-Net model was trained on 80% of these PET-MRI pairs and validated using 5-fold cross-validation. The models accuracy in reconstructed PET from MRI only was assessed using various intensity and noise metrics. ResultsThe model achieved a low voxel-wise mean squared error (0.0033 {+/-} 0.0010) across all folds and a median contrast-to-noise ratio of 0.0640 {+/-} 0.2500 when comparing true to reconstructed PET images. The synthesized PET images accurately replicated the spatial patterns observed in the original PET data. Additionally, the reconstruction accuracy was maintained even after spatial normalization. ConclusionThis study demonstrates that deep learning can accurately synthesize TSPO PET images from conventional, T1-weighted MRI. This approach could enable low-cost, noninvasive neuroinflammation imaging, expanding the clinical applicability of this imaging method.

neuroscience↗

bioRxiv2022

A translational MRI approach to validate acute axonal damage detection in multiple sclerosis

Axonal degeneration is a central pathological feature of neurodegenerative pathologies and is closely associated with irreversible clinical disability. Current noninvasive methods to detect axonal damage in vivo are limited in their specificity, clinical applicability, and lack of proper validation. We aimed to validate an MRI framework based on multicompartment modeling of the diffusion-weighted signal (AxCaliber) in rats in the presence of axonal pathology, achieved through injection of a neurotoxin damaging the neuronal terminal of axons. We then applied the same MRI protocol to map axonal integrity in the whole brain of multiple sclerosis relapsing-remitting patients and age-matched healthy controls, a pathology associated with a neurodegenerative component. AxCaliber is sensitive to microstructural axonal damage in rats, as demonstrated by a significant increase in the mean axonal caliber along the target tract, which correlated with the neurotoxin neurofilament staining. In humans, we uncovered a diffuse increase in mean axonal caliber in multiple sclerosis lesions and, importantly, in most areas of the normal-appearing white matter. Our results demonstrate that axonal diameter mapping is a sensitive and specific imaging biomarker able to link noninvasive imaging contrasts with the underlying biological substrate, supporting the key role of generalized axonal damage useful in diseases such as multiple sclerosis.

neuroscience↗

bioRxiv2021

Multimodal image fusion via deep generative models

Recently, it has become progressively more evident that classic diagnostic labels are unable to accurately and reliably describe the complexity and variability of several clinical phenotypes. This is particularly true for a broad range of neuropsychiatric illnesses such as depression and anxiety disorders or behavioural phenotypes such as aggression and antisocial personality. Patient heterogeneity can be better described and conceptualized by grouping individuals into novel categories, which are based on empirically-derived sections of intersecting continua that span both across and beyond traditional categorical borders. In this context, neuroimaging data carry a wealth of spatiotemporally resolved information about each patients brain. However, they are usually heavily collapsed a priori through procedures which are not learned as part of model training, and consequently not optimized for the downstream prediction task. This is due to the fact that every individual participant usually comes with multiple whole-brain 3D imaging modalities often accompanied by a deep genotypic and phenotypic characterization, hence posing formidable computational challenges. In this paper we design and validate a deep learning architecture based on generative models rooted in a modular approach and separable convolutional blocks (which result in a 20-fold decrease in parameter utilization) in order to a) fuse multiple 3D neuroimaging modalities on a voxel-wise level, b) efficiently convert them into informative latent embeddings through heavy dimensionality reduction, c) maintain excellent generalizability and minimal information loss. As proof of concept, we test our architecture on the well characterized Human Connectome Project database (n=974 healthy subjects), demonstrating that our latent embeddings can be clustered into easily separable subject strata which, in turn, map to extremely different phenotypical information (including organic, neuropsychological, personality variables) which was not included in the embedding creation process. The ability to extract meaningful and separable phenotypic information from brain images alone can aid in creating multi-dimensional biomarkers able to chart spatio-temporal trajectories which may correspond to different pathophysiological mechanisms unidentifiable to traditional data analysis approaches. In turn, this may be of aid in predicting disease evolution as well as drug response, hence supporting mechanistic disease understanding and also empowering clinical trials.

neuroscience↗

bioRxiv2020

A Deep Graph Neural Network Architecture for Modelling Spatio-temporal Dynamics in resting-stating functional MRI Data

Resting-state functional magnetic resonance imaging (rs-fMRI) has been successfully employed to understand the organisation of the human brain. For rs-fMRI analysis, the brain is typically parcellated into regions of interest (ROIs) and modelled as a graph where each ROI is a node and pairwise correlation between ROI blood-oxygen-level-dependent (BOLD) time series are edges. Recently, graph neural networks (GNNs) have seen a surge in popularity due to their successes in modelling unstructured relational data. The latest developments with GNNs, however, have not yet been fully exploited for the analysis of rs-fMRI data, particularly with regards to its spatio-temporal dynamics. Herein we present a novel deep neural network architecture, combining both GNNs and temporal convolutional networks (TCNs), which is able to learn from the spatial and temporal components of rs-fMRI data in an end-to-end fashion. In particular, this corresponds to intra-feature learning (i.e., learning temporal dynamics with TCNs) as well as inter-feature learning (i.e., leveraging spatial interactions between ROIs with GNNs). We evaluate our model with an ablation study using 35,159 samples from the UK Biobank rs-fMRI database. We also demonstrate explainability features of our architecture which map to realistic neurobiological insights. We hope our model could lay the groundwork for future deep learning architectures focused on leveraging the inherently and inextricably spatio-temporal nature of rs-fMRI data.

neuroscience↗

bioRxiv2019

Echo State Network models for nonlinear Granger causality

While Granger Causality (GC) has been often employed in network neuroscience, most GC applications are based on linear multivariate autoregressive (MVAR) models. However, real-life systems like biological networks exhibit notable non-linear behavior, hence undermining the validity of MVAR-based GC (MVAR-GC). Current nonlinear GC estimators only cater for additive nonlinearities or, alternatively, are based on recurrent neural networks (RNN) or Long short-term memory (LSTM) networks, which present considerable training difficulties and tailoring needs. We define a novel approach to estimating nonlinear, directed within-network interactions through a RNN class termed echo-state networks (ESN), where training is replaced by random initialization of an internal basis based on orthonormal matrices. We reformulate the GC framework in terms of ESN-based models, our ESN-based Granger Causality (ES-GC) estimator in a network of noisy Duffing oscillators, showing a net advantage of ES-GC in detecting nonlinear, causal links. We then explore the structure of ES-GC networks in the human brain employing functional MRI data from 1003 healthy subjects drawn from the human connectome project, demonstrating the existence of previously unknown directed within-brain interactions. ES-GC performs better than commonly used and recently developed GC approaches, making it a valuable tool for the analysis of e.g. multivariate biological networks.

neuroscience↗

bioRxiv2019

A machine learning tool for interpreting differences in cognition using brain features

Predicting variability in cognition traits is an attractive and challenging area of research, where different approaches and datasets have been implemented with mixed results. Some powerful Machine Learning algorithms employed before are difficult to interpret, while other algorithms are easy to interpret but might not be as powerful. To improve understanding of individual cognitive differences in humans, we make use of the most recent developments in Machine Learning in which powerful prediction models can be interpreted with confidence. We used neuroimaging data and a variety of behavioural, cognitive, affective and health measures from 905 people obtained from the Human Connectome Project (HCP). As a main contribution of this paper, we show how one could interpret the neuroanatomical basis of cognition, with recent methods which we believe are not yet fully explored in the field. By reducing neuroimages to a well characterised set of features generated from surface-based morphometry and cortical myelin estimates, we make the interpretation of such models easier as each feature is self-explanatory. The code used in this tool is available in a public repository: https://github.com/tjiagoM/interpreting-cognition-paper-2019

neuroscience↗

bioRxiv2018

A parameter-efficient deep learning approach to predict conversion from mild cognitive impairment to Alzheimer’s disease within three years

Some forms of mild cognitive impairment (MCI) can be the clinical precursor of severe dementia like Alzheimers disease (AD), while other types of MCI tend to remain stable over-time and do not progress to AD pathology. To choose an effective and personalized treatment for AD, we need to identify which MCI patients are at risk of developing AD and which are not.\n\nHere, we present a novel deep learning architecture, based on dual learning and an ad hoc layer for 3D separable convolutions, which aims at identifying those people with MCI who have a high likelihood of developing AD. Our deep learning procedures combine structural magnetic resonance imaging (MRI), demographic, neuropsychological, and APOe4 genotyping data as input measures. The most novel characteristics of our machine learning model compared to previous ones are as follows: 1) multi-tasking, in the sense that our deep learning model jointly learns to simultaneously predict both MCI to AD conversion, and AD vs healthy classification which facilitates the relevant feature extraction for prognostication; 2) the neural network classifier employs relatively few parameters compared to other deep learning architectures (we use ~550,000 network parameters, orders of magnitude lower than other network designs) without compromising network complexity and hence significantly limits data-overfitting; 3) both structural MRI images and warp field characteristics, which quantify the amount of volumetric change compared to the common template, were used as separate input streams to extract as much information as possible from the MRI data. All the analyses were performed on a subset of the Alzheimers Disease Neuroimaging Initiative (ADNI) database, for a total of n=785 participants (192 AD, 409 MCI, and184 healthy controls (HC)).\n\nWe found that the most predictive combination of inputs included the structural MRI images and the demographic, neuropsychological, and APOe4 data, while the warp field metric added little predictive value. We achieved an area under the ROC curve (AUC) of 0.925 with a 10-fold cross-validated accuracy of 86%, a sensitivity of 87.5% and specificity of 85% in classifying MCI patients who developed AD in three years time from those individuals showing stable MCI over the same time-period. To the best of our knowledge, this is the highest performance reported on a test set achieved in the literature using similar data. The same network provided an AUC of 1 and 100% accuracy, sensitivity and specificity when classifying NC from AD. We also demonstrated that our classification framework was robust to different co-registration templates and possibly irrelevant features / image sections.\n\nOur approach is flexible and can in principle integrate other imaging modalities, such as PET, and a more diverse group of clinical data. The convolutional framework is potentially applicable to any 3D image dataset and gives the flexibility to design a computer-aided diagnosis system targeting the prediction of any medical condition utilizing multi-modal imaging and tabular clinical data.

bioinformatics↗