Contrast-enhanced to non-contrast-enhanced image translation to exploit a clinical data warehouse of T1-weighted brain MRI.

BACKGROUND: Clinical data warehouses provide access to massive amounts of medical images, but these images are often heterogeneous. They can for instance include images acquired both with or without the injection of a gadolinium-based contrast agent. Harmonizing such data sets is thus fundamental to guarantee unbiased results, for example when performing differential diagnosis. Furthermore, classical neuroimaging software tools for feature extraction are typically applied only to images without gadolinium. The objective of this work is to evaluate how image translation can be useful to exploit a highly heterogeneous data set containing both contrast-enhanced and non-contrast-enhanced images from a clinical data warehouse. METHODS: We propose and compare different 3D U-Net and conditional GAN models to convert contrast-enhanced T1-weighted (T1ce) into non-contrast-enhanced (T1nce) brain MRI. These models were trained using 230 image pairs and tested on 77 image pairs from the clinical data warehouse of the Greater Paris area. RESULTS: Validation using standard image similarity measures demonstrated that the similarity between real and synthetic T1nce images was higher than between real T1nce and T1ce images for all the models compared. The best performing models were further validated on a segmentation task. We showed that tissue volumes extracted from synthetic T1nce images were closer to those of real T1nce images than volumes extracted from T1ce images. CONCLUSION: We showed that deep learning models initially developed with research quality data could synthesize T1nce from T1ce images of clinical quality and that reliable features could be extracted from the synthetic images, thus demonstrating the ability of such methods to help exploit a data set coming from a clinical data warehouse.

Contribution of MRI for the Early Diagnosis of Parkinsonism in Patients with Diagnostic Uncertainty.

BACKGROUND: International clinical criteria are the reference for the diagnosis of degenerative parkinsonism in clinical research, but they may lack sensitivity and specificity in the early stages. OBJECTIVES: To determine whether magnetic resonance imaging (MRI) analysis, through visual reading or machine-learning approaches, improves diagnostic accuracy compared with clinical diagnosis at an early stage in patients referred for suspected degenerative parkinsonism. MATERIALS: Patients with initial diagnostic uncertainty between Parkinson's disease (PD), progressive supranuclear palsy (PSP), and multisystem atrophy (MSA), with brain MRI performed at the initial visit (V1) and available 2-year follow-up (V2), were included. We evaluated the accuracy of the diagnosis established based on: (1) the international clinical diagnostic criteria for PD, PSP, and MSA at V1 ("Clin1"); (2) MRI visual reading blinded to the clinical diagnosis ("MRI"); (3) both MRI visual reading and clinical criteria at V1 ("MRI and Clin1"), and (4) a machine-learning algorithm ("Algorithm"). The gold standard diagnosis was established by expert consensus after a 2-year follow-up. RESULTS: We recruited 113 patients (53 with PD, 31 with PSP, and 29 with MSA). Considering the whole population, compared with clinical criteria at the initial visit ("Clin1": balanced accuracy, 66.2%), MRI visual reading showed a diagnostic gain of 14.3% ("MRI": 80.5%; P = 0.01), increasing to 19.2% when combined with the clinical diagnosis at the initial visit ("MRI and Clin1": 85.4%; P < 0.0001). The algorithm achieved a diagnostic gain of 9.9% ("Algorithm": 76.1%; P = 0.08). CONCLUSION: Our study shows the use of MRI analysis, whether by visual reading or machine-learning methods, for early differentiation of parkinsonism. © 2024 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Automatic motion artefact detection in brain T1-weighted magnetic resonance images from a clinical data warehouse using synthetic data.

Containing the medical data of millions of patients, clinical data warehouses (CDWs) represent a great opportunity to develop computational tools. Magnetic resonance images (MRIs) are particularly sensitive to patient movements during image acquisition, which will result in artefacts (blurring, ghosting and ringing) in the reconstructed image. As a result, a significant number of MRIs in CDWs are corrupted by these artefacts and may be unusable. Since their manual detection is impossible due to the large number of scans, it is necessary to develop tools to automatically exclude (or at least identify) images with motion in order to fully exploit CDWs. In this paper, we propose a novel transfer learning method from research to clinical data for the automatic detection of motion in 3D T1-weighted brain MRI. The method consists of two steps: a pre-training on research data using synthetic motion, followed by a fine-tuning step to generalise our pre-trained model to clinical data, relying on the labelling of 4045 images. The objectives were both (1) to be able to exclude images with severe motion, (2) to detect mild motion artefacts. Our approach achieved excellent accuracy for the first objective with a balanced accuracy nearly similar to that of the annotators (balanced accuracy>80 %). However, for the second objective, the performance was weaker and substantially lower than that of human raters. Overall, our framework will be useful to take advantage of CDWs in medical imaging and highlight the importance of a clinical validation of models trained on research data.

Proxy-analysis of the genetics of cognitive decline in Parkinson's disease through polygenic scores.

Cognitive decline is common in Parkinson's disease (PD) and its genetic risk factors are not well known to date, besides variants in the GBA and APOE genes. However, variation in complex traits is caused by numerous variants and is usually studied with genome-wide association studies (GWAS), requiring a large sample size, which is difficult to achieve for outcome measures in PD. Taking an alternative approach, we computed 100 polygenic scores (PGS) related to cognitive, dementia, stroke, and brain anatomical phenotypes and investigated their association with cognitive decline in six longitudinal cohorts. The analysis was adjusted for age, sex, genetic ancestry, follow-up duration, GBA and APOE status. Then, we meta-analyzed five of these cohorts, comprising a total of 1702 PD participants with 6156 visits, using the Montreal Cognitive Assessment as a cognitive outcome measure. After correction for multiple comparisons, we found four PGS significantly associated with cognitive decline: intelligence (p = 5.26e-13), cognitive performance (p = 1.46e-12), educational attainment (p = 8.52e-10), and reasoning (p = 3.58e-5). Survival analyses highlighted an offset of several years between the first and last quartiles of PGS, with significant differences for the PGS of cognitive performance (5 years) and educational attainment (7 years). In conclusion, we found four PGS associated with cognitive decline in PD, all associated with general cognitive phenotypes. This study highlights the common genetic factors between cognitive decline in PD and the general population, and the importance of the participant's cognitive reserve for cognitive outcome in PD.

Genomic loci influence patterns of structural covariance in the human brain.

Normal and pathologic neurobiological processes influence brain morphology in coordinated ways that give rise to patterns of structural covariance (PSC) across brain regions and individuals during brain aging and diseases. The genetic underpinnings of these patterns remain largely unknown. We apply a stochastic multivariate factorization method to a diverse population of 50,699 individuals (12 studies and 130 sites) and derive data-driven, multi-scale PSCs of regional brain size. PSCs were significantly correlated with 915 genomic loci in the discovery set, 617 of which are newly identified, and 72% were independently replicated. Key pathways influencing PSCs involve reelin signaling, apoptosis, neurogenesis, and appendage development, while pathways of breast cancer indicate potential interplays between brain metastasis and PSCs associated with neurodegeneration and dementia. Using support vector machines, multi-scale PSCs effectively derive imaging signatures of several brain diseases. Our results elucidate genetic and biological underpinnings that influence structural covariance patterns in the human brain.

Evaluation of MRI-based machine learning approaches for computer-aided diagnosis of dementia in a clinical data warehouse.

A variety of algorithms have been proposed for computer-aided diagnosis of dementia from anatomical brain MRI. These approaches achieve high accuracy when applied to research data sets but their performance on real-life clinical routine data has not been evaluated yet. The aim of this work was to study the performance of such approaches on clinical routine data, based on a hospital data warehouse, and to compare the results to those obtained on a research data set. The clinical data set was extracted from the hospital data warehouse of the Greater Paris area, which includes 39 different hospitals. The research set was composed of data from the Alzheimer's Disease Neuroimaging Initiative data set. In the clinical set, the population of interest was identified by exploiting the diagnostic codes from the 10th revision of the International Classification of Diseases that are assigned to each patient. We studied how the imbalance of the training sets, in terms of contrast agent injection and image quality, may bias the results. We demonstrated that computer-aided diagnosis performance was strongly biased upwards (over 17 percent points of balanced accuracy) by the confounders of image quality and contrast agent injection, a phenomenon known as the Clever Hans effect or shortcut learning. When these biases were removed, the performance was very poor. In any case, the performance was considerably lower than on the research data set. Our study highlights that there are still considerable challenges for translating dementia computer-aided diagnosis systems to clinical routine.

Temporo-basal sulcal connections: a manual annotation protocol and an investigation of sexual dimorphism and heritability.

The temporo-basal region of the human brain is composed of the collateral, the occipito-temporal, and the rhinal sulci. We manually rated (using a novel protocol) the connections between rhinal/collateral (RS-CS), collateral/occipito-temporal (CS-OTS) and rhinal/occipito-temporal (RS-OTS) sulci, using the MRI of nearly 3400 individuals including around 1000 twins. We reported both the associations between sulcal polymorphisms as well with a wide range of demographics (e.g. age, sex, handedness). Finally, we also estimated the heritability, and the genetic correlation between sulcal connections. We reported the frequency of the sulcal connections in the general population, which were hemisphere dependent. We found a sexual dimorphism of the connections, especially marked in the right hemisphere, with a CS-OTS connection more frequent in females (approximately 35-40% versus 20-25% in males) and an RS-CS connection more common in males (approximately 40-45% versus 25-30% in females). We confirmed associations between sulcal connections and characteristics of incomplete hippocampal inversion (IHI). We estimated the broad sense heritability to be 0.28-0.45 for RS-CS and CS-OTS connections, with hints of dominant contribution for the RS-CS connection. The connections appeared to share some of their genetic causing factors as indicated by strong genetic correlations. Heritability appeared much smaller for the (rarer) RS-OTS connection.

Automatic segmentation of the choroid plexuses: Method and validation in controls and patients with multiple sclerosis.

Choroid Plexuses (ChP) are structures located in the ventricles that produce the cerebrospinal fluid (CSF) in the central nervous system. They are also a key component of the blood-CSF barrier. Recent studies have described clinically relevant ChP volumetric changes in several neurological diseases including Alzheimer's, Parkinson's disease, and multiple sclerosis (MS). Therefore, a reliable and automated tool for ChP segmentation on images derived from magnetic resonance imaging (MRI) is a crucial need for large studies attempting to elucidate their role in neurological disorders. Here, we propose a novel automatic method for ChP segmentation in large imaging datasets. The approach is based on a 2-step 3D U-Net to keep preprocessing steps to a minimum for ease of use and to lower memory requirements. The models are trained and validated on a first research cohort including people with MS and healthy subjects. A second validation is also performed on a cohort of pre-symptomatic MS patients having acquired MRIs in routine clinical practice. Our method reaches an average Dice coefficient of 0.72 ± 0.01 with the ground truth and a volume correlation of 0.86 on the first cohort while outperforming FreeSurfer and FastSurfer-based ChP segmentations. On the dataset originating from clinical practice, the method reaches a Dice coefficient of 0.67 ± 0.01 (being close to the inter-rater agreement of 0.64 ± 0.02) and a volume correlation of 0.84. These results demonstrate that this is a suitable and robust method for the segmentation of the ChP both on research and clinical datasets.

Comparison of mean diffusivity, R2* relaxation rate and morphometric biomarkers for the clinical differentiation of parkinsonism.

INTRODUCTION: Quantitative biomarkers for clinical differentiation of parkinsonian syndromes are still lacking. Our aim was to evaluate the value of combining clinically feasible manual measurements of R2* relaxation rates and mean diffusivity (MD) in subcortical regions and brainstem morphometric measurements to improve the discrimination of parkinsonian syndromes. METHODS: Twenty-two healthy controls (HC), 25 patients with Parkinson's disease (PD), 19 with progressive supranuclear palsy (PSP) and 27 with multiple system atrophy (MSA, 21 with the parkinsonian variant -MSAp, 6 with the cerebellar variant -MSAc) were recruited. R2*, MD measurements and morphometric biomarkers including the midbrain to pons area ratio and the Magnetic Resonance Parkinsonism Index (MRPI) were compared between groups and their diagnostic performances were assessed. RESULTS: Morphometric biomarkers discriminated better patients with PSP (ratio: AUC 0.89, MRPI: AUC 0.89) and MSAc (ratio: AUC 0.82, MRPI: AUC 0.75) from other groups. R2* and MD measurements in the posterior putamen performed better in separating patients with MSAp from PD (R2*: AUC 0.89; MD: AUC 0.89). For the three-class classification "MSA vs PD vs PSP", the combination of MD and R2* measurements in the posterior putamen with morphometric biomarkers (AUC: 0.841) outperformed each marker separately. At the individual-level, there were seven discordances between imaging-based prediction and clinical diagnosis involving MSA. Using the new Movement Disorder Society criteria for the diagnosis of MSA, three of these seven patients were clinically reclassified as predicted by quantitative imaging. CONCLUSION: Combining R2* and MD measurements in the posterior putamen with morphometric biomarkers improves the discrimination of parkinsonism.

Online hard example mining vs. fixed oversampling strategy for segmentation of new multiple sclerosis lesions from longitudinal FLAIR MRI.

Detecting new lesions is a key aspect of the radiological follow-up of patients with Multiple Sclerosis (MS), leading to eventual changes in their therapeutics. This paper presents our contribution to the MSSEG-2 MICCAI 2021 challenge. The challenge is focused on the segmentation of new MS lesions using two consecutive Fluid Attenuated Inversion Recovery (FLAIR) Magnetic Resonance Imaging (MRI). In other words, considering longitudinal data composed of two time points as input, the aim is to segment the lesional areas, which are present only in the follow-up scan and not in the baseline. The backbone of our segmentation method is a 3D UNet applied patch-wise to the images, and in which, to take into account both time points, we simply concatenate the baseline and follow-up images along the channel axis before passing them to the 3D UNet. Our key methodological contribution is the use of online hard example mining to address the challenge of class imbalance. Indeed, there are very few voxels belonging to new lesions which makes training deep-learning models difficult. Instead of using handcrafted priors like brain masks or multi-stage methods, we experiment with a novel modification to online hard example mining (OHEM), where we use an exponential moving average (i.e., its weights are updated with momentum) of the 3D UNet to mine hard examples. Using a moving average instead of the raw model should allow smoothing of its predictions and allow it to give more consistent feedback for OHEM.

MicroRNA signatures in genetic frontotemporal dementia and amyotrophic lateral sclerosis.

OBJECTIVE: MicroRNAs are promising biomarkers of frontotemporal dementia (FTD) and amyotrophic lateral sclerosis (ALS), but discrepant results between studies have so far hampered their use in clinical trials. We aim to assess all previously identified circulating microRNA signatures as potential biomarkers of genetic FTD and/or ALS, using homogeneous, independent validation cohorts of C9orf72 and GRN mutation carriers. METHODS: 104 individuals carrying a C9orf72 or a GRN mutation, along with 31 controls, were recruited through the French research network on FTD/ALS. All subjects underwent blood sampling, from which circulating microRNAs were extracted. We measured differences in the expression levels of 65 microRNAs, selected from 15 published studies about FTD or ALS, between 31 controls, 17 C9orf72 presymptomatic subjects, and 29 C9orf72 patients. We also assessed differences in the expression levels of 30 microRNAs, selected from five studies about FTD, between 31 controls, 30 GRN presymptomatic subjects, and 28 GRN patients. RESULTS: More than half (35/65) of the selected microRNAs were differentially expressed in the C9orf72 cohort, while only a small proportion (5/30) of microRNAs were differentially expressed in the GRN cohort. In multivariate analyses, only individuals in the C9orf72 cohort could be adequately classified (ROC AUC up to 0.98 for controls versus presymptomatic subjects, 0.94 for controls versus patients, and 0.77 for presymptomatic subjects versus patients) with some of the signatures. INTERPRETATION: Our results suggest that previously identified microRNAs using sporadic or mixed cohorts of FTD and ALS patients could potentially serve as biomarkers of C9orf72-associated disease, but not GRN-associated disease.

The impact of aging on human brain network target controllability.

Understanding how few distributed areas can steer large-scale brain activity is a fundamental question that has practical implications, which range from inducing specific patterns of behavior to counteracting disease. Recent endeavors based on network controllability provided fresh insights into the potential ability of single regions to influence whole brain dynamics through the underlying structural connectome. However, controlling the entire brain activity is often unfeasible and might not always be necessary. The question whether single areas can control specific target subsystems remains crucial, albeit still poorly explored. Furthermore, the structure of the brain network exhibits progressive changes across the lifespan, but little is known about the possible consequences in the controllability properties. To address these questions, we adopted a novel target controllability approach that quantifies the centrality of brain nodes in controlling specific target anatomo-functional systems. We then studied such target control centrality in human connectomes obtained from healthy individuals aged from 5 to 85. Main results showed that the sensorimotor system has a high influencing capacity, but it is difficult for other areas to influence it. Furthermore, we reported that target control centrality varies with age and that temporal-parietal regions, whose cortical thinning is crucial in dementia-related diseases, exhibit lower values in older people. By simulating targeted attacks, such as those occurring in focal stroke, we showed that the ipsilesional hemisphere is the most affected one regardless of the damaged area. Notably, such degradation in target control centrality was more evident in younger people, thus supporting early-vulnerability hypotheses after stroke.

Disease Progression Score Estimation From Multimodal Imaging and MicroRNA Data Using Supervised Variational Autoencoders.

Frontotemporal dementia and amyotrophic lateral sclerosis are rare neurodegenerative diseases with no effective treatment. The development of biomarkers allowing an accurate assessment of disease progression is crucial for evaluating new therapies. Concretely, neuroimaging and transcriptomic (microRNA) data have been shown useful in tracking their progression. However, no single biomarker can accurately measure progression in these complex diseases. Additionally, large samples are not available for such rare disorders. It is thus essential to develop methods that can model disease progression by combining multiple biomarkers from small samples. In this paper, we propose a new framework for computing a disease progression score (DPS) from cross-sectional multimodal data. Specifically, we introduce a supervised multimodal variational autoencoder that can infer a meaningful latent space, where latent representations are placed along a disease trajectory. A score is computed by orthogonal projections onto this path. We evaluate our framework with multiple synthetic datasets and with a real dataset containing 14 patients, 40 presymptomatic genetic mutation carriers and 37 controls from the PREV-DEMALS study. There is no ground truth for the DPS in real-world scenarios, therefore we use the area under the ROC curve (AUC) as a proxy metric. Results with the synthetic datasets support this choice, since the higher the AUC, the more accurate the predicted simulated DPS. Experiments with the real dataset demonstrate better performance in comparison with state-of-the-art approaches. The proposed framework thus leverages cross-sectional multimodal datasets with small sample sizes to objectively measure disease progression, with potential application in clinical trials.

Machine Learning-Based Prediction of Impulse Control Disorders in Parkinson's Disease From Clinical and Genetic Data.

Goal: Impulse control disorders (ICDs) are frequent non-motor symptoms occurring during the course of Parkinson's disease (PD). The objective of this study was to estimate the predictability of the future occurrence of these disorders using longitudinal data, the first study using cross-validation and replication in an independent cohort. Methods: We used data from two longitudinal PD cohorts (training set: PPMI, Parkinson's Progression Markers Initiative; test set: DIGPD, Drug Interaction With Genes in Parkinson's Disease). We included 380 PD subjects from PPMI and 388 PD subjects from DIGPD, with at least two visits and with clinical and genetic data available, in our analyses. We trained three logistic regressions and a recurrent neural network to predict ICDs at the next visit using clinical risk factors and genetic variants previously associated with ICDs. We quantified performance using the area under the receiver operating characteristic curve (ROC AUC) and average precision. We compared these models to a trivial model predicting ICDs at the next visit with the status at the most recent visit. Results: The recurrent neural network (PPMI: 0.85 [0.80 - 0.90], DIGPD: 0.802 [0.78 - 0.83]) was the only model to be significantly better than the trivial model (PPMI: ROC AUC = 0.75 [0.69 - 0.81]; DIGPD: 0.78 [0.75 - 0.80]) on both cohorts. We showed that ICDs in PD can be predicted with better accuracy with a recurrent neural network model than a trivial model. The improvement in terms of ROC AUC was higher on PPMI than on DIGPD data, but not clinically relevant in both cohorts. Conclusions: Our results indicate that machine learning methods are potentially useful for predicting ICDs, but further works are required to reach clinical relevance.

Parsimonious model for mass-univariate vertexwise analysis.

Purpose: Covariance between gray-matter measurements can reflect structural or functional brain networks though it has also been shown to be influenced by confounding factors (e.g., age, head size, and scanner), which could lead to lower mapping precision (increased size of associated clusters) and create distal false positives associations in mass-univariate vertexwise analyses. Approach: We evaluated this concern by performing state-of-the-art mass-univariate analyses (general linear model, GLM) on traits simulated from real vertex-wise gray matter data (including cortical and subcortical thickness and surface area). We contrasted the results with those from linear mixed models (LMMs), which have been shown to overcome similar issues in omics association studies. Results: We showed that when performed on a large sample ( N = 8662 , UK Biobank), GLMs yielded greatly inflated false positive rate (cluster false discovery rate > 0.6 ). We showed that LMMs resulted in more parsimonious results: smaller clusters and reduced false positive rate but at a cost of increased computation. Next, we performed mass-univariate association analyses on five real UKB traits (age, sex, BMI, fluid intelligence, and smoking status) and LMM yielded fewer and more localized associations. We identified 19 significant clusters displaying small associations with age, sex, and BMI, which suggest a complex architecture of at least dozens of associated areas with those phenotypes. Conclusions: The published literature could contain a large proportion of redundant (possibly confounded) associations that are largely prevented using LMMs. The parsimony of LMMs results from controlling for the joint effect of all vertices, which prevents local and distal redundant associations from reaching significance.

Association of APOE-Independent Alzheimer Disease Polygenic Risk Score With Brain Amyloid Deposition in Asymptomatic Older Adults.

BACKGROUND AND OBJECTIVES: Brain amyloid deposition, a major risk factor for Alzheimer's disease (AD), is currently estimated by measuring cerebrospinal fluid or plasma amyloid peptide levels, or by positron-emission tomography imaging. Assessing genetic risks relating to amyloid deposition before any accumulation has occurred would allow for earlier intervention in persons at increased risk for developing AD. Previous work linking amyloid burden and genetic risk relied almost exclusively on APOE, a major AD genetic risk factor. Here, we ask whether a polygenic risk score (PRS) that incorporates an optimized list of common variants linked to AD and excludes APOE is associated with brain amyloid load in cognitively unimpaired elderly adults. METHODS: We included 291 elderly asymptomatic participants from the INveStIGation of AlzHeimer's PredicTors (INSIGHT-preAD) cohort who underwent amyloid imaging, including 83 amyloid-positive (+) participants. We used an Alzheimer's (A) PRS composed of 33 AD risk variants excluding APOE, and selected the 17 variants that showed the strongest association with amyloid positivity to define an optimized (oA) PRS. Participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI) study [228 participants, 90 amyloid (+)] were tested as a validation cohort. Finally, 2,300 AD patients and 6,994 controls from the European Alzheimer's Disease Initiative (EADI) were evaluated. RESULTS: A-PRS was not significantly associated with amyloid burden in the INSIGHT or ADNI cohorts with or without correction for APOE genotype. However, oA-PRS was significantly associated with amyloid status independently of APOE adjustment (INSIGHT OR: 5.26 [1.71-16.88]; ADNI OR: 3.38 [1.02-11.63]). Interestingly, oA-PRS accurately discriminated amyloid (+) and (-) APOE ε4 carriers (INSIGHT OR: 181.6 [7.53-10,674.6]; ADNI OR: 44.94 [3.03-1,277]). A-PRS and oA-PRS showed a significant association with disease status in the EADI cohort (OR: 1.68 [1.53-1.85] and 2.06 [1.73-2.45] respectively). Genes assigned to oA-PRS variants were enriched in ontologies related to Aß metabolism and deposition. DISCUSSION: PRSs relying on AD genetic risk factors excluding APOE may improve risk prediction for brain amyloid, allowing stratification of cognitively unimpaired individuals at risk of AD independent of their APOE status.

ClinicaDL: An open-source deep learning software for reproducible neuroimaging processing.

BACKGROUND AND OBJECTIVE: As deep learning faces a reproducibility crisis and studies on deep learning applied to neuroimaging are contaminated by methodological flaws, there is an urgent need to provide a safe environment for deep learning users to help them avoid common pitfalls that will bias and discredit their results. Several tools have been proposed to help deep learning users design their framework for neuroimaging data sets. Software overview: We present here ClinicaDL, one of these software tools. ClinicaDL interacts with BIDS, a standard format in the neuroimaging field, and its derivatives, so it can be used with a large variety of data sets. Moreover, it checks the absence of data leakage when inferring the results of new data with trained networks, and saves all necessary information to guarantee the reproducibility of results. The combination of ClinicaDL and its companion project Clinica allows performing an end-to-end neuroimaging analysis, from the download of raw data sets to the interpretation of trained networks, including neuroimaging preprocessing, quality check, label definition, architecture search, and network training and evaluation. CONCLUSIONS: We implemented ClinicaDL to bring answers to three common issues encountered by deep learning users who are not always familiar with neuroimaging data: (1) the format and preprocessing of neuroimaging data sets, (2) the contamination of the evaluation procedure by data leakage and (3) a lack of reproducibility. We hope that its use by researchers will allow producing more reliable and thus valuable scientific studies in our field.

Multilevel Survival Modeling With Structured Penalties for Disease Prediction From Imaging Genetics Data.

This paper introduces a framework for disease prediction from multimodal genetic and imaging data. We propose a multilevel survival model which allows predicting the time of occurrence of a future disease state in patients initially exhibiting mild symptoms. This new multilevel setting allows modeling the interactions between genetic and imaging variables. This is in contrast with classical additive models which treat all modalities in the same manner and can result in undesirable elimination of specific modalities when their contributions are unbalanced. Moreover, the use of a survival model allows overcoming the limitations of previous approaches based on classification which consider a fixed time frame. Furthermore, we introduce specific penalties taking into account the structure of the different types of data, such as a group lasso penalty over the genetic modality and a l2-penalty over the imaging modality. Finally, we propose a fast optimization algorithm, based on a proximal gradient method. The approach was applied to the prediction of Alzheimer's disease (AD) among patients with mild cognitive impairment (MCI) based on genetic (single nucleotide polymorphisms - SNP) and imaging (anatomical MRI measures) data from the ADNI database. The experiments demonstrate the effectiveness of the method for predicting the time of conversion to AD. It revealed how genetic variants and brain imaging alterations interact in the prediction of future disease status. The approach is generic and could potentially be useful for the prediction of other diseases.

Automatic quality control of brain T1-weighted magnetic resonance images for a clinical data warehouse.

Many studies on machine learning (ML) for computer-aided diagnosis have so far been mostly restricted to high-quality research data. Clinical data warehouses, gathering routine examinations from hospitals, offer great promises for training and validation of ML models in a realistic setting. However, the use of such clinical data warehouses requires quality control (QC) tools. Visual QC by experts is time-consuming and does not scale to large datasets. In this paper, we propose a convolutional neural network (CNN) for the automatic QC of 3D T1-weighted brain MRI for a large heterogeneous clinical data warehouse. To that purpose, we used the data warehouse of the hospitals of the Greater Paris area (Assistance Publique-Hôpitaux de Paris [AP-HP]). Specifically, the objectives were: 1) to identify images which are not proper T1-weighted brain MRIs; 2) to identify acquisitions for which gadolinium was injected; 3) to rate the overall image quality. We used 5000 images for training and validation and a separate set of 500 images for testing. In order to train/validate the CNN, the data were annotated by two trained raters according to a visual QC protocol that we specifically designed for application in the setting of a data warehouse. For objectives 1 and 2, our approach achieved excellent accuracy (balanced accuracy and F1-score >90%), similar to the human raters. For objective 3, the performance was good but substantially lower than that of human raters. Nevertheless, the automatic approach accurately identified (balanced accuracy and F1-score >80%) low quality images, which would typically need to be excluded. Overall, our approach shall be useful for exploiting hospital data warehouses in medical image computing.