• Publicationsclose
• Other research productsclose
• Open Access close
• English close

Recent advances in MRI have led to the creation of large datasets. With the increase in data volume, it has become difficult to locate previous scans of the same patient within these datasets (a process known as re-identification). To address this issue, we propose an AI-powered medical imaging retrieval framework called DeepBrainPrint, which is designed to retrieve brain MRI scans of the same patient. Our framework is a semi-self-supervised contrastive deep learning approach with three main innovations. First, we use a combination of self-supervised and supervised paradigms to create an effective brain fingerprint from MRI scans that can be used for real-time image retrieval. Second, we use a special weighting function to guide the training and improve model convergence. Third, we introduce new imaging transformations to improve retrieval robustness in the presence of intensity variations (i.e. different scan contrasts), and to account for age and disease progression in patients. We tested DeepBrainPrint on a large dataset of T1-weighted brain MRIs from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and on a synthetic dataset designed to evaluate retrieval performance with different image modalities. Our results show that DeepBrainPrint outperforms previous methods, including simple similarity metrics and more advanced contrastive deep learning frameworks.

The brain-age gap is one of the most investigated risk markers for brain changes across disorders. While the field is progressing towards large-scale models, recently incorporating uncertainty estimates, no model to date provides the single-subject risk assessment capability essential for clinical application. In order to enable the clinical use of brain-age as a biomarker, we here combine uncertainty-aware deep Neural Networks with conformal prediction theory. This approach provides statistical guarantees with respect to single-subject uncertainty estimates and allows for the calculation of an individual's probability for accelerated brain-aging. Building on this, we show empirically in a sample of N=16,794 participants that 1. a lower or comparable error as state-of-the-art, large-scale brain-age models, 2. the statistical guarantees regarding single-subject uncertainty estimation indeed hold for every participant, and 3. that the higher individual probabilities of accelerated brain-aging derived from our model are associated with Alzheimer's Disease, Bipolar Disorder and Major Depressive Disorder. Comment: arXiv admin note: text overlap with arXiv:2107.07977

In recent years, the rapid development of artificial intelligence has promoted the widespread application of convolutional neural networks (CNNs) in neuroimaging analysis. Although three-dimensional (3D) CNNs can utilize the spatial information in 3D volumes, there are still some challenges related to high-dimensional features and potential overfitting issues. To overcome these problems, patch-based CNNs have been used, which are beneficial for model generalization. However, it is unclear how the choice of a patchwise sampling strategy affects the performance of the Alzheimer’s Disease (AD) classification. To this end, the present work investigates the impact of a patchwise sampling strategy for 3D CNN based AD classification. A 3D framework cascaded by two-stage subnetworks was used for AD classification. The patch-level subnetworks learned feature representations from local image patches, and the subject-level subnetwork combined discriminative feature representations from all patch-level subnetworks to generate a classification score at the subject level. Experiments were conducted to determine the effect of patch partitioning methods, the effect of patch size, and interactions between patch size and training set size for AD classification. With the same data size and identical network structure, the 3D CNN model trained with 48 × 48 × 48 cubic image patches showed the best performance in AD classification (ACC = 89.6%). The model trained with hippocampus-centered, region of interest (ROI)-based image patches showed suboptimal performance. If the pathological features are concentrated only in some regions affected by the disease, the empirically predefined ROI patches might be the right choice. The better performance of cubic image patches compared with cuboidal image patches is likely related to the pathological distribution of AD. The image patch size and training sample size together have a complex influence on the performance of the classification. The size of the image patches should be determined based on the size of the training sample to compensate for noisy labels and the problem of the curse of dimensionality. The conclusions of the present study can serve as a reference for the researchers who wish to develop a superior 3D patch-based CNN model with an appropriate patch sampling strategy.

This paper considers canonical correlation analysis for two longitudinal variables that are possibly sampled at different time resolutions with irregular grids. We modeled trajectories of the multivariate variables using random effects and found the most correlated sets of linear combinations in the latent space. Our numerical simulations showed that the longitudinal canonical correlation analysis effectively recovers underlying correlation patterns between two high-dimensional longitudinal data sets. We applied the proposed LCCA to data from the Alzheimer's Disease Neuroimaging Initiative and identified the longitudinal profiles of morphological brain changes and amyloid cumulation. Comment: 24 pages, 16 figures

The hippocampus is one of the most studied neuroanatomical structures due to its involvement in attention, learning, and memory as well as its atrophy in ageing, neurological, and psychiatric diseases. Hippocampal shape changes, however, are complex and cannot be fully characterized by a single summary metric such as hippocampal volume as determined from MR images. In this work, we propose an automated, geometry-based approach for the unfolding, point-wise correspondence, and local analysis of hippocampal shape features such as thickness and curvature. Starting from an automated segmentation of hippocampal subfields, we create a 3D tetrahedral mesh model as well as a 3D intrinsic coordinate system of the hippocampal body. From this coordinate system, we derive local curvature and thickness estimates as well as a 2D sheet for hippocampal unfolding. We evaluate the performance of our algorithm with a series of experiments to quantify neurodegenerative changes in Mild Cognitive Impairment and Alzheimer's disease dementia. We find that hippocampal thickness estimates detect known differences between clinical groups and can determine the location of these effects on the hippocampal sheet. Further, thickness estimates improve classification of clinical groups and cognitively unimpaired controls when added as an additional predictor. Comparable results are obtained with different datasets and segmentation algorithms. Taken together, we replicate canonical findings on hippocampal volume/shape changes in dementia, extend them by gaining insight into their spatial localization on the hippocampal sheet, and provide additional, complementary information beyond traditional measures. We provide a new set of sensitive processing and analysis tools for the analysis of hippocampal geometry that allows comparisons across studies without relying on image registration or requiring manual intervention. Comment: Updated to journal publication

Covariate adjustment and methods of incorporating historical data in randomized clinical trials (RCTs) each provide opportunities to increase trial power. We unite these approaches for the analysis of RCTs with binary outcomes based on the Cochran-Mantel-Haenszel (CMH) test for marginal risk ratio (RR). In PROCOVA-CMH, subjects are stratified on a single prognostic covariate reflective of their predicted outcome on the control treatment (e.g. placebo). This prognostic score is generated based on baseline covariates through a model trained on historical data. We propose two closed-form prospective estimators for the asymptotic sampling variance of the log RR that rely only on values obtainable from observed historical outcomes and the prognostic model. Importantly, these estimators can be used to inform sample size during trial planning. PROCOVA-CMH demonstrates type I error control and appropriate asymptotic coverage for valid inference. Like other covariate adjustment methods, PROCOVA-CMH can reduce the variance of the treatment effect estimate when compared to an unadjusted (unstratified) CMH analysis. In addition to statistical methods, simulations and a case study in Alzheimer's Disease are given to demonstrate performance. Results show that PROCOVA-CMH can provide a gain in power, which can be used to conduct smaller trials. Comment: 18 pages, 11 tables, appendix

Die vorliegende Arbeit beschäftigt sich mit der Herausforderung der Entwicklung statistischer Formmodelle, welche die inhärente nichteuklidische Struktur (anatomischer) Formvariation berücksichtigen, gleichzeitig eine effiziente, numerisch robuste Verarbeitung erlauben und zusätzlich möglichst viel Invarianz unter euklidischen Bewegungen der verwendeten Daten bieten. Dazu schlagen wir einen kontinuierlichen und physikalisch motivierten Formenraum basierend auf Deformationsgradienten vor. Wir verfolgen zwei verschiedene Ansätze, um auf diesem eine Riemannsche Struktur und damit ein statistisches Formmodell zu etablieren. (1) Wir entwickeln ein Modell für die differentiellen Koordinaten als Elemente in GL(3)+. Zu diesem Zweck adaptieren wir den Begriff des biinvarianten Mittelwerts bezüglich eines affinen Zusammenhanges auf GL(3)+ und führen Statistik zweiter Ordnung basierend auf einer Familie maximal invarianter, d.h. GL(3)+-links- und O(3)-rechts-invarianter, Riemannscher Metriken durch. (2) Wir versehen die differentiellen Koordinaten mit einer nichteuklidischen Struktur, die angelehnt an die Lie-Produktgruppe aus Streckungen und Rotationen ist. Diese lässt eine biinvariante Metrik und damit eine konsistente Analyse mittels mannigfaltigkeitswertiger Statistik im Riemannschen Rahmen zu. Die vorliegende Arbeit präsentiert überdies die fundamentalen Koordinaten, eine neue Formrepräsentation basierend auf diskreten Flächenfundamentalformen, welche auf natürliche Weise invariant unter euklidischen Bewegungen ist. Wir versehen diese Repräsentation mit einer Lie-Gruppenstruktur, die eine biinvariante Metrik und somit Riemannsche Statistik erlaubt. Darüber hinaus entwickeln wir einen einfachen, effizienten, robusten, sowie akkuraten (d.h. ohne Rückgriff auf Modellapproximationen) Löser für die Rückabbildung von den Koordinaten zur Form im Raum. Neben der statistischen Formanalyse erlaubt der beschriebene Ansatz auch Anwendungen in der Geometrieverarbeitung, insbesondere zur quasi-isometrischen Oberflächenverflachung. Der letzte Abschnitt der Arbeit befasst sich mit der Entwicklung kontinuierlicher formbasierter Erkrankungsstratifikationen, um die Krankheitsbewertung über die aktuelle klinische Praxis ordinaler Bewertungssysteme hinaus zu objektivieren. Hierzu entwickeln wir den geodesic B-score in gekrümmten Formenräumen zur Bewertung von Kniegelenksarthrose als Generalisierung des euklidischen B-scores. In diesem Rahmen beschreiben wir eine Newton-Typ-Fixpunktiteration zur Bestimmung der Projektion auf Geodätische im Formenraum. Als Anwendung zeigen wir, dass der geodätische B-score gegenüber seinem euklidischen Gegenstück eine verbesserte Vorhersageleistung hinsichtlich der Risikobewertung bezüglich einer totalen Kniearthroplastie besitzt. In this work, we address the challenge of developing statistical shape models that account for the non-Euclidean nature inherent to (anatomical) shape variation and at the same time offer fast, numerically robust processing and as much invariance as possible regarding translation and rotation, i.e. Euclidean motion. With the aim of doing that we formulate a continuous and physically motivated notion of shape space based on deformation gradients. We follow two different tracks endowing this differential representation with a Riemannian structure to establish a statistical shape model. (1) We derive a model based on differential coordinates as elements in GL(3)+. To this end, we adapt the notion of bi-invariant means employing an affine connection structure on GL(3)+. Furthermore, we perform second-order statistics based on a family of Riemannian metrics providing the most possible invariance, viz. GL(3)+-left-invariance and O(3)-right-invariance. (2) We endow the differential coordinates with a non-Euclidean structure, that stems from a product Lie group of stretches and rotations. This structure admits a bi-invariant metric and thus allows for a consistent analysis via manifold-valued Riemannian statistics. This work further presents a novel shape representation based on discrete fundamental forms that is naturally invariant under Euclidean motion, namely the fundamental coordinates. We endow this representation with a Lie group structure that admits bi-invariant metrics and therefore allows for consistent analysis using manifold-valued statistics based on the Riemannian framework. Furthermore, we derive a simple, efficient, robust, yet accurate (i.e. without resorting to model approximations) solver for the inverse problem that allows for interactive applications. Beyond statistical shape modeling the proposed framework is amenable for surface processing such as quasi-isometric flattening. Additionally, the last part of the thesis aims on shape-based, continuous disease stratification to provide means that objectify disease assessment over the current clinical practice of ordinal grading systems. Therefore, we derive the geodesic B-score, a generalization of the of the Euclidean B-score, in order to assess knee osteoarthritis. In this context we present a Newton-type fixed point iteration for projection onto geodesics in shape space. On the application side, we show that the derived geodesic B-score features, in comparison to its Euclidean counterpart, an improved predictive performance on assessing the risk of total knee replacement surgery.

Background: In recent years, predicting and modeling the progression of Alzheimer’s disease (AD) based on neuropsychological tests has become increasingly appealing in AD research.Objective: In this study, we aimed to predict the neuropsychological scores and investigate the non-linear progression trend of the cognitive declines based on multimodal neuroimaging data.Methods: We utilized unimodal/bimodal neuroimaging measures and a non-linear regression method (based on artificial neural networks) to predict the neuropsychological scores in a large number of subjects (n = 1143), including healthy controls (HC) and patients with mild cognitive impairment non-converter (MCI-NC), mild cognitive impairment converter (MCI-C), and AD. We predicted two neuropsychological scores, i.e., the clinical dementia rating sum of boxes (CDRSB) and Alzheimer’s disease assessment scale cognitive 13 (ADAS13), based on structural magnetic resonance imaging (sMRI) and positron emission tomography (PET) biomarkers.Results: Our results revealed that volumes of the entorhinal cortex and hippocampus and the average fluorodeoxyglucose (FDG)-PET of the angular gyrus, temporal gyrus, and posterior cingulate outperform other neuroimaging features in predicting ADAS13 and CDRSB scores. Compared to a unimodal approach, our results showed that a bimodal approach of integrating the top two neuroimaging features (i.e., the entorhinal volume and the average FDG of the angular gyrus, temporal gyrus, and posterior cingulate) increased the prediction performance of ADAS13 and CDRSB scores in the converting and stable stages of MCI and AD. Finally, a non-linear AD progression trend was modeled to describe the cognitive decline based on neuroimaging biomarkers in different stages of AD.Conclusion: Findings in this study show an association between neuropsychological scores and sMRI and FDG-PET biomarkers from normal aging to severe AD.

Objectives: Neuropsychological tests are an important basis for the memory impairment diagnosis in Alzheimer’s disease (AD). However, multiple memory tests might be conflicting within-subjects and lead to uncertain diagnoses in some cases. This study proposed a framework to diagnose the uncertain cases of memory impairment.Methods: We collected 2,386 samples including AD, mild cognitive impairment (MCI), and cognitive normal (CN) using 18F-fluorodeoxyglucose positron emission tomography (FDG-PET) and three different neuropsychological tests (Mini-Mental State Examination, Alzheimer’s Disease Assessment Scale-Cognitive Subscale, and Clinical Dementia Rating) from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). A deep learning (DL) framework using FDG-PET was proposed to diagnose uncertain memory impairment cases that were conflicting between tests. Subsequent ANOVA, chi-squared, and t-test were used to explain the potential causes of uncertain cases.Results: For certain cases in the testing set, the proposed DL framework outperformed other methods with 95.65% accuracy. For the uncertain cases, its positive diagnoses had a significant (p < 0.001) worse decline in memory function than negative diagnoses in a longitudinal study of 40 months on average. In the memory-impaired group, uncertain cases were mainly explained by an AD metabolism pattern but mild in extent (p < 0.05). In the healthy group, uncertain cases were mainly explained by a non-energetic mental state (p < 0.001) measured using a global deterioration scale (GDS), with a significant depression-related metabolism pattern detected (p < 0.05).Conclusion: A DL framework for diagnosing uncertain cases of memory impairment is proposed. Proved by longitudinal tracing of its diagnoses, it showed clinical validity and had application potential. Its valid diagnoses also provided evidence and explanation of uncertain cases based on the neurodegeneration and depression mental state.

The recent proliferation of medical data, such as genetics and electronic health records (EHR), offers new opportunities to find novel predictors of health outcomes. Presented with a large set of candidate features, interest often lies in selecting the ones most likely to be predictive of an outcome for further study such that the goal is to control the false discovery rate (FDR) at a specified level. Knockoff filtering is an innovative strategy for FDR-controlled feature selection. But, existing knockoff methods make strong distributional assumptions that hinder their applicability to real world data. We propose Bayesian models for generating high quality knockoff copies that utilize available knowledge about the data structure, thus improving the resolution of prognostic features. Applications to two feature sets are considered: those with categorical and/or continuous variables possibly having a population substructure, such as in EHR; and those with microbiome features having a compositional constraint and phylogenetic relatedness. Through simulations and real data applications, these methods are shown to identify important features with good FDR control and power.