Prediction of Severe Acute Pancreatitis at a Very Early Stage of the Disease Using Artificial Intelligence Techniques, Without Laboratory Data or Imaging Tests: The PANCREATIA Study.

OBJECTIVE: To evaluate machine learning models' performance in predicting acute pancreatitis severity using early-stage variables while excluding laboratory and imaging tests. SUMMARY BACKGROUND DATA: Severe acute pancreatitis (SAP) affects approximately 20% of acute pancreatitis (AP) patients and is associated with high mortality rates. Accurate early prediction of SAP and in-hospital mortality is crucial for effective management. Traditional scores such as APACHE-II and BISAP are complex and require laboratory tests, while early predictive models are lacking. Machine learning (ML) has shown promising results in predictive modelling, potentially outperforming traditional methods. METHODS: We analysed data from a prospective database of AP patients admitted to Vall d'Hebron Hospital from November 2015 to January 2022. Inclusion criteria were adults diagnosed with AP according to the 2012 Atlanta classification. Data included basal characteristics, current medication, and vital signs. We developed machine learning models to predict SAP, in-hospital mortality, and intensive care unit (ICU) admission. The modelling process included two stages: Stage 0, which used basal characteristics and medication, and Stage 1, which included data from Stage 0 and vital signs. RESULTS: Out of 634 cases, 594 were analysed. The Stage 0 model showed AUC values of 0.698 for mortality, 0.721 for ICU admission, and 0.707 for persistent organ failure. The Stage 1 model improved performance with AUC values of 0.849 for mortality, 0.786 for ICU admission, and 0.783 for persistent organ failure. The models demonstrated comparable or superior performance to APACHE-II and BISAP scores. CONCLUSIONS: The ML models showed good predictive capacity for SAP, ICU admission, and mortality using early-stage data without laboratory or imaging tests. This approach could revolutionise AP patients' initial triage and management, providing a personalised prediction method based on early clinical data.

Enhanced Artificial Intelligence Methods for Liver Steatosis Assessment Using Machine Learning and Color Image Processing: Liver Color Project.

BACKGROUND: The use of livers with significant steatosis is associated with worse transplantation outcomes. Brain death donor liver acceptance is mostly based on subjective surgeon assessment of liver appearance, since steatotic livers acquire a yellowish tone. The aim of this study was to develop a rapid, robust, accurate, and cost-effective method to assess liver steatosis. METHODS: From June 1, 2018, to November 30, 2023, photographs and tru-cut needle biopsies were taken from adult brain death donor livers at a single university hospital for the study. All the liver photographs were taken by smartphones then color calibrated, segmented, and divided into patches. Color and texture features were then extracted and used as input, and the machine learning method was applied. This is a collaborative project between Vall d'Hebron University Hospital and Barcelona MedTech, Pompeu Fabra University, and is referred to as LiverColor. RESULTS: A total of 192 livers (362 photographs and 7240 patches) were included. When setting a macrosteatosis threshold of 30%, the best results were obtained using the random forest classifier, achieving an AUROC = 0.74, with 85% accuracy. CONCLUSION: Machine learning coupled with liver texture and color analysis of photographs taken with smartphones provides excellent accuracy for determining liver steatosis.

MC-RVAE: Multi-channel recurrent variational autoencoder for multimodal Alzheimer's disease progression modelling.

The progression of neurodegenerative diseases, such as Alzheimer's Disease, is the result of complex mechanisms interacting across multiple spatial and temporal scales. Understanding and predicting the longitudinal course of the disease requires harnessing the variability across different data modalities and time, which is extremely challenging. In this paper, we propose a model based on recurrent variational autoencoders that is able to capture cross-channel interactions between different modalities and model temporal information. These are achieved thanks to its multi-channel architecture and its shared latent variational space, parametrized with a recurrent neural network. We evaluate our model on both synthetic and real longitudinal datasets, the latter including imaging and non-imaging data, with N=897 subjects. Results show that our multi-channel recurrent variational autoencoder outperforms a set of baselines (KNN, random forest, and group factor analysis) for the task of reconstructing missing modalities, reducing the mean absolute error by 5% (w.r.t. the best baseline) for both subcortical volumes and cortical thickness. Our model is robust to missing features within each modality and is able to generate realistic synthetic imaging biomarkers trajectories from cognitive scores.

A Riemannian approach to predicting brain function from the structural connectome.

Ongoing brain function is largely determined by the underlying wiring of the brain, but the specific rules governing this relationship remain unknown. Emerging literature has suggested that functional interactions between brain regions emerge from the structural connections through mono- as well as polysynaptic mechanisms. Here, we propose a novel approach based on diffusion maps and Riemannian optimization to emulate this dynamic mechanism in the form of random walks on the structural connectome and predict functional interactions as a weighted combination of these random walks. Our proposed approach was evaluated in two different cohorts of healthy adults (Human Connectome Project, HCP; Microstructure-Informed Connectomics, MICs). Our approach outperformed existing approaches and showed that performance plateaus approximately around the third random walk. At macroscale, we found that the largest number of walks was required in nodes of the default mode and frontoparietal networks, underscoring an increasing relevance of polysynaptic communication mechanisms in transmodal cortical networks compared to primary and unimodal systems.

Nonlinear interaction between APOE ε4 allele load and age in the hippocampal surface of cognitively intact individuals.

The ε4 allele of the gene Apolipoprotein E is the major genetic risk factor for Alzheimer's Disease. APOE ε4 has been associated with changes in brain structure in cognitively impaired and unimpaired subjects, including atrophy of the hippocampus, which is one of the brain structures that is early affected by AD. In this work we analyzed the impact of APOE ε4 gene dose and its association with age, on hippocampal shape assessed with multivariate surface analysis, in a ε4-enriched cohort of n = 479 cognitively healthy individuals. Furthermore, we sought to replicate our findings on an independent dataset of n = 969 individuals covering the entire AD spectrum. We segmented the hippocampus of the subjects with a multi-atlas-based approach, obtaining high-dimensional meshes that can be analyzed in a multivariate way. We analyzed the effects of different factors including APOE, sex, and age (in both cohorts) as well as clinical diagnosis on the local 3D hippocampal surface changes. We found specific regions on the hippocampal surface where the effect is modulated by significant APOE ε4 linear and quadratic interactions with age. We compared between APOE and diagnosis effects from both cohorts, finding similarities between APOE ε4 and AD effects on specific regions, and suggesting that age may modulate the effect of APOE ε4 and AD in a similar way.

Fetal cortical surface atlas parcellation based on growth patterns.

Defining anatomically and functionally meaningful parcellation maps on cortical surface atlases is of great importance in surface-based neuroimaging analysis. The conventional cortical parcellation maps are typically defined based on anatomical cortical folding landmarks in adult surface atlases. However, they are not suitable for fetal brain studies, due to dramatic differences in brain size, shape, and properties between adults and fetuses. To address this issue, we propose a novel data-driven method for parcellation of fetal cortical surface atlases into distinct regions based on the dynamic "growth patterns" of cortical properties (e.g., surface area) from a population of fetuses. Our motivation is that the growth patterns of cortical properties indicate the underlying rapid changes of microstructures, which determine the molecular and functional principles of the cortex. Thus, growth patterns are well suitable for defining distinct cortical regions in development, structure, and function. To comprehensively capture the similarities of cortical growth patterns among vertices, we construct two complementary similarity matrices. One is directly based on the growth trajectories of vertices, and the other is based on the correlation profiles of vertices' growth trajectories in relation to a set of reference points. Then, we nonlinearly fuse these two similarity matrices into a single one, which can better capture both their common and complementary information than by simply averaging them. Finally, based on this fused similarity matrix, we perform spectral clustering to divide the fetal cortical surface atlases into distinct regions. By applying our method on 25 normal fetuses from 26 to 29 gestational weeks, we construct age-specific fetal cortical surface atlases equipped with biologically meaningful parcellation maps based on cortical growth patterns. Importantly, our generated parcellation maps reveal spatially contiguous, hierarchical and bilaterally relatively symmetric patterns of fetal cortical surface development.

Toward the automatic quantification of in utero brain development in 3D structural MRI: A review.

Investigating the human brain in utero is important for researchers and clinicians seeking to understand early neurodevelopmental processes. With the advent of fast magnetic resonance imaging (MRI) techniques and the development of motion correction algorithms to obtain high-quality 3D images of the fetal brain, it is now possible to gain more insight into the ongoing maturational processes in the brain. In this article, we present a review of the major building blocks of the pipeline toward performing quantitative analysis of in vivo MRI of the developing brain and its potential applications in clinical settings. The review focuses on T1- and T2-weighted modalities, and covers state of the art methodologies involved in each step of the pipeline, in particular, 3D volume reconstruction, spatio-temporal modeling of the developing brain, segmentation, quantification techniques, and clinical applications. Hum Brain Mapp 38:2772-2787, 2017. © 2017 Wiley Periodicals, Inc.

Diffusion maps for multimodal registration.

Multimodal image registration is a difficult task, due to the significant intensity variations between the images. A common approach is to use sophisticated similarity measures, such as mutual information, that are robust to those intensity variations. However, these similarity measures are computationally expensive and, moreover, often fail to capture the geometry and the associated dynamics linked with the images. Another approach is the transformation of the images into a common space where modalities can be directly compared. Within this approach, we propose to register multimodal images by using diffusion maps to describe the geometric and spectral properties of the data. Through diffusion maps, the multimodal data is transformed into a new set of canonical coordinates that reflect its geometry uniformly across modalities, so that meaningful correspondences can be established between them. Images in this new representation can then be registered using a simple Euclidean distance as a similarity measure. Registration accuracy was evaluated on both real and simulated brain images with known ground-truth for both rigid and non-rigid registration. Results showed that the proposed approach achieved higher accuracy than the conventional approach using mutual information.

Accuracy and repeatability of fetal facial measurements in 3D ultrasound: A longitudinal study.

OBJECTIVE: Fetal face measurements in prenatal ultrasound can aid in identifying craniofacial abnormalities in the developing fetus. However, the accuracy and reliability of ultrasound measurements can be affected by factors such as fetal position, image quality, and the sonographer's expertise. This study assesses the accuracy and reliability of fetal facial measurements in prenatal ultrasound. Additionally, the temporal evolution of measurements is studied, comparing prenatal and postnatal measurements. METHODS: Three different experts located up to 23 facial landmarks in 49 prenatal 3D ultrasound scans from normal Caucasian fetuses at weeks 20, 26, and 35 of gestation. Intra- and inter-observer variability was obtained. Postnatal facial measurements were also obtained at 15 days and 1 month postpartum. RESULTS: Most facial landmarks exhibited low errors, with overall intra- and inter-observer errors of 1.01 mm and 1.60 mm, respectively. Landmarks on the nose were found to be the most reliable, while the most challenging ones were those located on the ears and eyes. Overall, scans obtained at 26 weeks of gestation presented the best trade-off between observer variability and landmark visibility. The temporal evolution of the measurements revealed that the lower face area had the highest rate of growth throughout the latest stages of pregnancy. CONCLUSIONS: Craniofacial landmarks can be evaluated using 3D fetal ultrasound, especially those located on the nose, mouth, and chin. Despite its limitations, this study provides valuable insights into prenatal and postnatal biometric changes over time, which could aid in developing predictive models for postnatal measurements based on prenatal data.

Unveiling interactions between intervertebral disc morphologies and mechanical behavior through personalized finite element modeling.

Introduction: Intervertebral Disc (IVD) Degeneration (IDD) is a significant health concern, potentially influenced by mechanotransduction. However, the relationship between the IVD phenotypes and mechanical behavior has not been thoroughly explored in local morphologies where IDD originates. This work unveils the interplays among morphological and mechanical features potentially relevant to IDD through Abaqus UMAT simulations. Methods: A groundbreaking automated method is introduced to transform a calibrated, structured IVD finite element (FE) model into 169 patient-personalized (PP) models through a mesh morphing process. Our approach accurately replicates the real shapes of the patient's Annulus Fibrosus (AF) and Nucleus Pulposus (NP) while maintaining the same topology for all models. Using segmented magnetic resonance images from the former project MySpine, 169 models with structured hexahedral meshes were created employing the Bayesian Coherent Point Drift++ technique, generating a unique cohort of PP FE models under the Disc4All initiative. Machine learning methods, including Linear Regression, Support Vector Regression, and eXtreme Gradient Boosting Regression, were used to explore correlations between IVD morphology and mechanics. Results: We achieved PP models with AF and NP similarity scores of 92.06\% and 92.10\% compared to the segmented images. The models maintained good quality and integrity of the mesh. The cartilage endplate (CEP) shape was represented at the IVD-vertebra interfaces, ensuring personalized meshes. Validation of the constitutive model against literature data showed a minor relative error of 5.20%. Discussion: Analysis revealed the influential impact of local morphologies on indirect mechanotransduction responses, highlighting the roles of heights, sagittal areas, and volumes. While the maximum principal stress was influenced by morphologies such as heights, the disc's ellipticity influenced the minimum principal stress. Results suggest the CEPs are not influenced by their local morphologies but by those of the AF and NP. The generated free-access repository of individual disc characteristics is anticipated to be a valuable resource for the scientific community with a broad application spectrum.

Characterization of clinical data for patient stratification in moderate osteoarthritis with support vector machines, regulatory network models, and verification against osteoarthritis Initiative data.

Knee osteoarthritis (OA) diagnosis is based on symptoms, assessed through questionnaires such as the WOMAC. However, the inconsistency of pain recording and the discrepancy between joint phenotype and symptoms highlight the need for objective biomarkers in knee OA diagnosis. To this end, we study relationships among clinical and molecular data in a cohort of women (n = 51) with Kellgren-Lawrence grade 2-3 knee OA through a Support Vector Machine (SVM) and a regulation network model. Clinical descriptors (i.e., pain catastrophism, depression, functionality, joint pain, rigidity, sensitization and synovitis) are used to classify patients. A Youden's test is performed for each classifier to determine optimal binarization thresholds for the descriptors. Thresholds are tested against patient stratification according to baseline WOMAC data from the Osteoarthritis Initiative, and the mean accuracy is 0.97. For our cohort, the data used as SVM inputs are knee OA descriptors, synovial fluid proteomic measurements (n = 25), and transcription factor activation obtained from regulatory network model stimulated with the synovial fluid measurements. The relative weights after classification reflect input importance. The performance of each classifier is evaluated through ROC-AUC analysis. The best classifier with clinical data is pain catastrophism (AUC = 0.9), highly influenced by funcionality and pain sensetization, suggesting that kinesophobia is involved in pain perception. With synovial fluid proteins used as input, leptin strongly influences every classifier, suggesting the importance of low-grade inflammation. When transcription factors are used, the mean AUC is limited to 0.608, which can be related to the pleomorphic behaviour of osteoarthritic chondrocytes. Nevertheless, funcionality has an AUC of 0.7 with a decisive importance of FOXO downregulation. Though larger and longitudinal cohorts are needed, this unique combination of SVM and regulatory network model shall help to stratify knee OA patients more objectively.

LiverColor: An Artificial Intelligence Platform for Liver Graft Assessment.

Hepatic steatosis, characterized by excess fat in the liver, is the main reason for discarding livers intended for transplantation due to its association with increased postoperative complications. The current gold standard for evaluating hepatic steatosis is liver biopsy, which, despite its accuracy, is invasive, costly, slow, and not always feasible during liver procurement. Consequently, surgeons often rely on subjective visual assessments based on the liver's colour and texture, which are prone to errors and heavily depend on the surgeon's experience. The aim of this study was to develop and validate a simple, rapid, and accurate method for detecting steatosis in donor livers to improve the decision-making process during liver procurement. We developed LiverColor, a co-designed software platform that integrates image analysis and machine learning to classify a liver graft into valid or non-valid according to its steatosis level. We utilized an in-house dataset of 192 cases to develop and validate the classification models. Colour and texture features were extracted from liver photographs, and graft classification was performed using supervised machine learning techniques (random forests and support vector machine). The performance of the algorithm was compared against biopsy results and surgeons' classifications. Usability was also assessed in simulated and real clinical settings using the Mobile Health App Usability Questionnaire. The predictive models demonstrated an area under the receiver operating characteristic curve of 0.82, with an accuracy of 85%, significantly surpassing the accuracy of visual inspections by surgeons. Experienced surgeons rated the platform positively, appreciating not only the hepatic steatosis assessment but also the dashboarding functionalities for summarising and displaying procurement-related data. The results indicate that image analysis coupled with machine learning can effectively and safely identify valid livers during procurement. LiverColor has the potential to enhance the accuracy and efficiency of liver assessments, reducing the reliance on subjective visual inspections and improving transplantation outcomes.

Automatic and interpretable prediction of the site of origin in outflow tract ventricular arrhythmias: machine learning integrating electrocardiograms and clinical data.

The treatment of outflow tract ventricular arrhythmias (OTVA) through radiofrequency ablation requires the precise identification of the site of origin (SOO). Pinpointing the SOO enhances the likelihood of a successful procedure, reducing intervention times and recurrence rates. Current clinical methods to identify the SOO are based on qualitative analysis of pre-operative electrocardiograms (ECG), heavily relying on physician's expertise. Although computational models and machine learning (ML) approaches have been proposed to assist OTVA procedures, they either consume substantial time, lack interpretability or do not use clinical information. Here, we propose an alternative strategy for automatically predicting the ventricular origin of OTVA patients using ML. Our objective was to classify ventricular (left/right) origin in the outflow tracts (LVOT and RVOT, respectively), integrating ECG and clinical data from each patient. Extending beyond differentiating ventricle origin, we explored specific SOO characterization. Utilizing four databases, we also trained supervised learning models on the QRS complexes of the ECGs, clinical data, and their combinations. The best model achieved an accuracy of 89%, highlighting the significance of precordial leads V1-V4, especially in the R/S transition and initiation of the QRS complex in V2. Unsupervised analysis revealed that some origins tended to group closer than others, e.g., right coronary cusp (RCC) with a less sparse group than the aortic cusp origins, suggesting identifiable patterns for specific SOOs.

Altered Temporal Dynamics of Resting-State Functional Magnetic Resonance Imaging in Adolescent-Onset First-Episode Psychosis.

BACKGROUND: Dynamic functional connectivity (dFC) alterations have been reported in patients with adult-onset and chronic psychosis. We sought to examine whether such abnormalities were also observed in patients with first episode, adolescent-onset psychosis (AOP), in order to rule out potential effects of chronicity and protracted antipsychotic treatment exposure. AOP has been suggested to have less diagnostic specificity compared to psychosis with onset in adulthood and occurs during a period of neurodevelopmental changes in brain functional connections. STUDY DESIGN: Seventy-nine patients with first episode, AOP (36 patients with schizophrenia-spectrum disorder, SSD; and 43 with affective psychotic disorder, AF) and 54 healthy controls (HC), aged 10 to 17 years were included. Participants underwent clinical and cognitive assessments and resting-state functional magnetic resonance imaging. Graph-based measures were used to analyze temporal trajectories of dFC, which were compared between patients with SSD, AF, and HC. Within patients, we also tested associations between dFC parameters and clinical variables. STUDY RESULTS: Patients with SSD temporally visited the different connectivity states in a less efficient way (reduced global efficiency), visiting fewer nodes (larger temporal modularity, and increased immobility), with a reduction in the metabolic expenditure (cost and leap size), relative to AF and HC (effect sizes: Cohen's D, ranging 0.54 to.91). In youth with AF, these parameters did not differ compared to HC. Connectivity measures were not associated with clinical severity, intelligence, cannabis use, or dose of antipsychotic medication. CONCLUSIONS: dFC measures hold potential towards the development of brain-based biomarkers characterizing adolescent-onset SSD.

Network-based modelling of mechano-inflammatory chondrocyte regulation in early osteoarthritis.

Osteoarthritis (OA) is a debilitating joint disease characterized by articular cartilage degradation, inflammation and pain. An extensive range of in vivo and in vitro studies evidences that mechanical loads induce changes in chondrocyte gene expression, through a process known as mechanotransduction. It involves cascades of complex molecular interactions that convert physical signals into cellular response(s) that favor either chondroprotection or cartilage destruction. Systematic representations of those interactions can positively inform early strategies for OA management, and dynamic modelling allows semi-quantitative representations of the steady states of complex biological system according to imposed initial conditions. Yet, mechanotransduction is rarely integrated. Hence, a novel mechano-sensitive network-based model is proposed, in the form of a continuous dynamical system: an interactome of a set of 118 nodes, i.e., mechano-sensitive cellular receptors, second messengers, transcription factors and proteins, related among each other through a specific topology of 358 directed edges is developed. Results show that under physio-osmotic initial conditions, an anabolic state is reached, whereas initial perturbations caused by pro-inflammatory and injurious mechanical loads leads to a catabolic profile of node expression. More specifically, healthy chondrocyte markers (Sox9 and CITED2) are fully expressed under physio-osmotic conditions, and reduced under inflammation, or injurious loadings. In contrast, NF-κB and Runx2, characteristic of an osteoarthritic chondrocyte, become activated under inflammation or excessive loading regimes. A literature-based evaluation shows that the model can replicate 94% of the experiments tested. Sensitivity analysis based on a factorial design of a treatment shows that inflammation has the strongest influence on chondrocyte metabolism, along with a significant deleterious effect of static compressive loads. At the same time, anti-inflammatory therapies appear as the most promising ones, though the restoration of structural protein production seems to remain a major challenge even in beneficial mechanical environments. The newly developed mechano-sensitive network model for chondrocyte activity reveals a unique potential to reflect load-induced chondroprotection or articular cartilage degradation in different mechano-chemical-environments.

Memory-aware curriculum federated learning for breast cancer classification.

BACKGROUND AND OBJECTIVE: For early breast cancer detection, regular screening with mammography imaging is recommended. Routine examinations result in datasets with a predominant amount of negative samples. The limited representativeness of positive cases can be problematic for learning Computer-Aided Diagnosis (CAD) systems. Collecting data from multiple institutions is a potential solution to mitigate this problem. Recently, federated learning has emerged as an effective tool for collaborative learning. In this setting, local models perform computation on their private data to update the global model. The order and the frequency of local updates influence the final global model. In the context of federated adversarial learning to improve multi-site breast cancer classification, we investigate the role of the order in which samples are locally presented to the optimizers. METHODS: We define a novel memory-aware curriculum learning method for the federated setting. We aim to improve the consistency of the local models penalizing inconsistent predictions, i.e., forgotten samples. Our curriculum controls the order of the training samples prioritizing those that are forgotten after the deployment of the global model. Our approach is combined with unsupervised domain adaptation to deal with domain shift while preserving data privacy. RESULTS: Two classification metrics: area under the receiver operating characteristic curve (ROC-AUC) and area under the curve for the precision-recall curve (PR-AUC) are used to evaluate the performance of the proposed method. Our method is evaluated with three clinical datasets from different vendors. An ablation study showed the improvement of each component of our method. The AUC and PR-AUC are improved on average by 5% and 6%, respectively, compared to the conventional federated setting. CONCLUSIONS: We demonstrated the benefits of curriculum learning for the first time in a federated setting. Our results verified the effectiveness of the memory-aware curriculum federated learning for the multi-site breast cancer classification. Our code is publicly available at: https://github.com/ameliajimenez/curriculum-federated-learning.

An automatic pipeline for atlas-based fetal and neonatal brain segmentation and analysis.

BACKGROUND AND OBJECTIVE: The automatic segmentation of perinatal brain structures in magnetic resonance imaging (MRI) is of utmost importance for the study of brain growth and related complications. While different methods exist for adult and pediatric MRI data, there is a lack for automatic tools for the analysis of perinatal imaging. METHODS: In this work, a new pipeline for fetal and neonatal segmentation has been developed. We also report the creation of two new fetal atlases, and their use within the pipeline for atlas-based segmentation, based on novel registration methods. The pipeline is also able to extract cortical and pial surfaces and compute features, such as curvature, local gyrification index, sulcal depth, and thickness. RESULTS: Results show that the introduction of the new templates together with our segmentation strategy leads to accurate results when compared to expert annotations, as well as better performances when compared to a reference pipeline (developing Human Connectome Project (dHCP)), for both early and late-onset fetal brains. CONCLUSIONS: These findings show the potential of the presented atlases and the whole pipeline for application in both fetal, neonatal, and longitudinal studies, which could lead to dramatic improvements in the understanding of perinatal brain development.

Effects of Mediterranean diet or mindfulness-based stress reduction on fetal and neonatal brain development: a secondary analysis of a randomized clinical trial.

BACKGROUND: Maternal suboptimal nutrition and high stress levels are associated with adverse fetal and infant neurodevelopment. OBJECTIVE: This study aimed to investigate if structured lifestyle interventions involving a Mediterranean diet or mindfulness-based stress reduction during pregnancy are associated with differences in fetal and neonatal brain development. STUDY DESIGN: This was a secondary analysis of the randomized clinical trial Improving Mothers for a Better Prenatal Care Trial Barcelona that was conducted in Barcelona, Spain, from 2017 to 2020. Participants with singleton pregnancies were randomly allocated into 3 groups, namely Mediterranean diet intervention, stress reduction program, or usual care. Participants in the Mediterranean diet group received monthly individual sessions and free provision of extra-virgin olive oil and walnuts. Pregnant women in the stress reduction group underwent an 8-week mindfulness-based stress reduction program adapted for pregnancy. Magnetic resonance imaging of 90 fetal brains was performed at 36 to 39 weeks of gestation and the Neonatal Neurobehavioral Assessment Scale was completed for 692 newborns at 1 to 3 months. Fetal outcomes were the total brain volume and lobular or regional volumes obtained from a 3-dimensional reconstruction and semiautomatic segmentation of magnetic resonance images. Neonatal outcomes were the 6 clusters scores of the Neonatal Neurobehavioral Assessment Scale. Multiple regression analyses were conducted to assess the association between the interventions and the fetal and neonatal outcomes. RESULTS: When compared with the usual care group, the offspring exposed to a maternal Mediterranean diet had a larger total fetal brain volume (mean, 284.11 cm3; standard deviation, 23.92 cm3 vs 294.01 cm3; standard deviation, 26.29 cm3; P=.04), corpus callosum (mean, 1.16 cm3; standard deviation, 0.19 cm3 vs 1.26 cm3; standard deviation, 0.22 cm3; P=.03), and right frontal lobe (44.20; standard deviation, 4.09 cm3 vs 46.60; standard deviation, 4.69 cm3; P=.02) volumes based on magnetic resonance imaging measures and higher scores in the Neonatal Neurobehavioral Assessment Scale clusters of autonomic stability (mean, 7.4; standard deviation, 0.9 vs 7.6; standard deviation, 0.7; P=.04), social interaction (mean, 7.5; standard deviation, 1.5 vs 7.8; standard deviation, 1.3; P=.03), and range of state (mean, 4.3; standard deviation, 1.3 vs 4.5; standard deviation, 1.0; P=.04). When compared with the usual care group, offspring from the stress reduction group had larger fetal left anterior cingulate gyri volume (1.63; standard deviation, 0.32 m3 vs 1.79; standard deviation, 0.30 cm3; P=.03) based on magnetic resonance imaging and higher scores in the Neonatal Neurobehavioral Assessment Scale for regulation of state (mean, 6.0; standard deviation, 1.8 vs 6.5; standard deviation, 1.5; P<.01). CONCLUSION: Maternal structured lifestyle interventions involving the promotion of a Mediterranean diet or stress reduction during pregnancy were associated with changes in fetal and neonatal brain development.

Riemannian Geometry of Functional Connectivity Matrices for Multi-Site Attention-Deficit/Hyperactivity Disorder Data Harmonization.

The use of multi-site datasets in neuroimaging provides neuroscientists with more statistical power to perform their analyses. However, it has been shown that the imaging-site introduces variability in the data that cannot be attributed to biological sources. In this work, we show that functional connectivity matrices derived from resting-state multi-site data contain a significant imaging-site bias. To this aim, we exploited the fact that functional connectivity matrices belong to the manifold of symmetric positive-definite (SPD) matrices, making it possible to operate on them with Riemannian geometry. We hereby propose a geometry-aware harmonization approach, Rigid Log-Euclidean Translation, that accounts for this site bias. Moreover, we adapted other Riemannian-geometric methods designed for other domain adaptation tasks and compared them to our proposal. Based on our results, Rigid Log-Euclidean Translation of multi-site functional connectivity matrices seems to be among the studied methods the most suitable in a clinical setting. This represents an advance with respect to previous functional connectivity data harmonization approaches, which do not respect the geometric constraints imposed by the underlying structure of the manifold. In particular, when applying our proposed method to data from the ADHD-200 dataset, a multi-site dataset built for the study of attention-deficit/hyperactivity disorder, we obtained results that display a remarkable correlation with established pathophysiological findings and, therefore, represent a substantial improvement when compared to the non-harmonization analysis. Thus, we present evidence supporting that harmonization should be extended to other functional neuroimaging datasets and provide a simple geometric method to address it.

Prediction of Lung Nodule Progression with an Uncertainty-Aware Hierarchical Probabilistic Network.

Predicting whether a lung nodule will grow, remain stable or regress over time, especially early in its follow-up, would help doctors prescribe personalized treatments and better surgical planning. However, the multifactorial nature of lung tumour progression hampers the identification of growth patterns. In this work, we propose a deep hierarchical generative and probabilistic network that, given an initial image of the nodule, predicts whether it will grow, quantifies its future size and provides its expected semantic appearance at a future time. Unlike previous solutions, our approach also estimates the uncertainty in the predictions from the intrinsic noise in medical images and the inter-observer variability in the annotations. The evaluation of this method on an independent test set reported a future tumour growth size mean absolute error of 1.74 mm, a nodule segmentation Dice's coefficient of 78% and a tumour growth accuracy of 84% on predictions made up to 24 months ahead. Due to the lack of similar methods for providing future lung tumour growth predictions, along with their associated uncertainty, we adapted equivalent deterministic and alternative generative networks (i.e., probabilistic U-Net, Bayesian test dropout and Pix2Pix). Our method outperformed all these methods, corroborating the adequacy of our approach.