We address the prediction of non-imaging variables based on structural brain connectivity derived from diffusion magnetic resonance images, using graph-based machine learning. We predict age and the mini-mental state examination (MMSE) score as examples of a demographic and a clinical variable. We propose a machine-learning model inspired by graph convolutional networks (GCNs), which takes a brain connectivity graph as input and processes the data separately through a parallel GCN mechanism with multiple branches. The novelty of our work lies in the model architecture, especially the Connectivity Attention Block, which learns an embedding representation of brain graphs while providing graph-level attention. We show experiments on publicly available datasets of PREVENT-AD and OASIS3. We validate our model by comparing it to existing methods and via ablations. This quantifies the degree to which the connectome varies depending on the task, which is important for improving our understanding of health and disease across the population. The proposed model generally demonstrates higher performance especially for age prediction compared to the existing machine-learning algorithms we tested, including classical methods and (graph and non-graph) deep learning.

Due to the significant amount of time and expertise needed for manual segmentation of the brain cortex from magnetic resonance imaging (MRI) data, there is a substantial need for efficient and accurate algorithms to replace the need for human involvement. In this work, we explore the capabilities of Graph Neural Networks (GNNs) to segment the brain surface based on structural brain connectivity. We train three different GNN architectures, the Graph Convolutional Network (GCN), the Graph Attention Network (GAT), and the Graph U-Net, and evaluate their performances when trained on silver-standard cortical region labels created by FreeSurfer. We take a multimodal approach to brain segmentation by examining the influence of the structural connectivity values inferred from diffusion MRI (dMRI) in addition to using values from structural MRI (sMRI). Our results demonstrate the utility of GNN models, particularly the GAT architecture, which achieved Dice scores competitive to those reported in the literature with non-graph methods. Additionally, structural connectivity derived from dMRI revealed significant value in improving automatic segmentation, as models trained on combined attributes from dMRI and sMRI outperformed those trained only on sMRI. Finally, we compared the GNN-based and the FreeSurfer segmentations in their ability to predict demographic/clinical data, where neither of the two approaches was statistically significantly superior to the other.

Graph neural networks are powerful tools that enable deep learning on non-Euclidean data structures like graphs, point clouds, and meshes. They leverage the connectivity of data points and can even benefit learning tasks on data, which is not naturally graph-structured -like point clouds. In these cases, the graph structure needs to be determined from the dataset, which adds a significant challenge to the learning process. This opens up a multitude of design choices for creating suitable graph structures, which have a substantial impact on the success of the graph learning task. However, so far no concrete guidance for choosing the most appropriate graph construction is available, not only due to the large variety of methods out there but also because of its strong connection to the dataset at hand. In medicine, for example, a large variety of different data types complicates the selection of graph construction methods even more. We therefore summarise the current state-of-the-art graph construction methods, especially for medical data. In this work, we introduce a categorisation scheme for graph types and graph construction methods. We identify two main strands of graph construction: static and adaptive methods, discuss their advantages and disadvantages, and formulate recommendations for choosing a suitable graph construction method. We furthermore discuss how a created graph structure can be assessed and to what degree it supports graph learning. We hope to support medical research with graph deep learning with this work by elucidating the wide variety of graph construction methods.

We tackle classification based on brain connectivity derived from diffusion magnetic resonance images. We propose a machine-learning model inspired by graph convolutional networks (GCNs), which takes a brain connectivity input graph and processes the data separately through a parallel GCN mechanism with multiple heads. The proposed network is a simple design that employs different heads involving graph convolutions focused on edges and nodes, capturing representations from the input data thoroughly. To test the ability of our model to extract complementary and representative features from brain connectivity data, we chose the task of sex classification. This quantifies the degree to which the connectome varies depending on the sex, which is important for improving our understanding of health and disease in both sexes. We show experiments on two publicly available datasets: PREVENT-AD (347 subjects) and OASIS3 (771 subjects). The proposed model demonstrates the highest performance compared to the existing machine-learning algorithms we tested, including classical methods and (graph and non-graph) deep learning. We provide a detailed analysis of each component of our model.

This study highlights the importance of conducting comprehensive model inspection as part of comparative performance analyses. Here, we investigate the effect of modelling choices on the feature learning characteristics of graph neural networks applied to a brain shape classification task. Specifically, we analyse the effect of using parameter-efficient, shared graph convolutional submodels compared to structure-specific, non-shared submodels. Further, we assess the effect of mesh registration as part of the data harmonisation pipeline. We find substantial differences in the feature embeddings at different layers of the models. Our results highlight that test accuracy alone is insufficient to identify important model characteristics such as encoded biases related to data source or potentially non-discriminative features learned in submodels. Our model inspection framework offers a valuable tool for practitioners to better understand performance characteristics of deep learning models in medical imaging.

Neural networks are increasingly finding their way into the realm of graphs and modeling relationships between features. Concurrently graph neural network explanation approaches are being invented to uncover relationships between the nodes of the graphs. However, there is a disparity between the existing attribution methods, and it is unclear which attribution to trust. Therefore research has introduced evaluation experiments that assess them from different perspectives. In this work, we assess attribution methods from a perspective not previously explored in the graph domain: retraining. The core idea is to retrain the network on important (or not important) relationships as identified by the attributions and evaluate how networks can generalize based on these relationships. We reformulate the retraining framework to sidestep issues lurking in the previous formulation and propose guidelines for correct analysis. We run our analysis on four state-of-the-art GNN attribution methods and five synthetic and real-world graph classification datasets. The analysis reveals that attributions perform variably depending on the dataset and the network. Most importantly, we observe that the famous GNNExplainer performs similarly to an arbitrary designation of edge importance. The study concludes that the retraining evaluation cannot be used as a generalized benchmark and recommends it as a toolset to evaluate attributions on a specifically addressed network, dataset, and sparsity.

The standard of care for brain tumors is maximal safe surgical resection. Neuronavigation augments the surgeon's ability to achieve this but loses validity as surgery progresses due to brain shift. Moreover, gliomas are often indistinguishable from surrounding healthy brain tissue. Intraoperative magnetic resonance imaging (iMRI) and ultrasound (iUS) help visualize the tumor and brain shift. iUS is faster and easier to incorporate into surgical workflows but offers a lower contrast between tumorous and healthy tissues than iMRI. With the success of data-hungry Artificial Intelligence algorithms in medical image analysis, the benefits of sharing well-curated data cannot be overstated. To this end, we provide the largest publicly available MRI and iUS database of surgically treated brain tumors, including gliomas (n=92), metastases (n=11), and others (n=11). This collection contains 369 preoperative MRI series, 320 3D iUS series, 301 iMRI series, and 356 segmentations collected from 114 consecutive patients at a single institution. This database is expected to help brain shift and image analysis research and neurosurgical training in interpreting iUS and iMRI.

Pre-training has shown success in different areas of machine learning, such as Computer Vision, Natural Language Processing (NLP), and medical imaging. However, it has not been fully explored for clinical data analysis. An immense amount of clinical records are recorded, but still, data and labels can be scarce for data collected in small hospitals or dealing with rare diseases. In such scenarios, pre-training on a larger set of unlabeled clinical data could improve performance. In this paper, we propose novel unsupervised pre-training techniques designed for heterogeneous, multi-modal clinical data for patient outcome prediction inspired by masked language modeling (MLM), by leveraging graph deep learning over population graphs. To this end, we further propose a graph-transformer-based network, designed to handle heterogeneous clinical data. By combining masking-based pre-training with a transformer-based network, we translate the success of masking-based pre-training in other domains to heterogeneous clinical data. We show the benefit of our pre-training method in a self-supervised and a transfer learning setting, utilizing three medical datasets TADPOLE, MIMIC-III, and a Sepsis Prediction Dataset. We find that our proposed pre-training methods help in modeling the data at a patient and population level and improve performance in different fine-tuning tasks on all datasets.

Graph deep learning has recently emerged as a powerful ML concept allowing to generalize successful deep neural architectures to non-euclidean structured data. Such methods have shown promising results on a broad spectrum of applications ranging from social science, biomedicine, and particle physics to computer vision, graphics, and chemistry. One of the limitations of the majority of current graph neural network architectures is that they are often restricted to the transductive setting and rely on the assumption that the underlying graph is known and fixed. Often, this assumption is not true since the graph may be noisy, or partially and even completely unknown. In such cases, it would be helpful to infer the graph directly from the data, especially in inductive settings where some nodes were not present in the graph at training time. Furthermore, learning a graph may become an end in itself, as the inferred structure may provide complementary insights next to the downstream task. In this paper, we introduce Differentiable Graph Module (DGM), a learnable function that predicts edge probabilities in the graph which are optimal for the downstream task. DGM can be combined with convolutional graph neural network layers and trained in an end-to-end fashion. We provide an extensive evaluation of applications from the domains of healthcare (disease prediction), brain imaging (age prediction), computer graphics (3D point cloud segmentation), and computer vision (zero-shot learning). We show that our model provides a significant improvement over baselines both in transductive and inductive settings and achieves state-of-the-art results.

Graphs are a powerful tool for representing and analyzing unstructured, non-Euclidean data ubiquitous in the healthcare domain. Two prominent examples are molecule property prediction and brain connectome analysis. Importantly, recent works have shown that considering relationships between input data samples has a positive regularizing effect on the downstream task in healthcare applications. These relationships are naturally modeled by a (possibly unknown) graph structure between input samples. In this work, we propose Graph-in-Graph (GiG), a neural network architecture for protein classification and brain imaging applications that exploits the graph representation of the input data samples and their latent relation. We assume an initially unknown latent-graph structure between graph-valued input data and propose to learn a parametric model for message passing within and across input graph samples, end-to-end along with the latent structure connecting the input graphs. Further, we introduce a Node Degree Distribution Loss (NDDL) that regularizes the predicted latent relationships structure. This regularization can significantly improve the downstream task. Moreover, the obtained latent graph can represent patient population models or networks of molecule clusters, providing a level of interpretability and knowledge discovery in the input domain, which is of particular value in healthcare.

Interpretability in Graph Convolutional Networks (GCNs) has been explored to some extent in general in computer vision; yet, in the medical domain, it requires further examination. Most of the interpretability approaches for GCNs, especially in the medical domain, focus on interpreting the output of the model in a post-hoc fashion. In this paper, we propose an interpretable attention module (IAM) that explains the relevance of the input features to the classification task on a GNN Model. The model uses these interpretations to improve its performance. In a clinical scenario, such a model can assist the clinical experts in better decision-making for diagnosis and treatment planning. The main novelty lies in the IAM, which directly operates on input features. IAM learns the attention for each feature based on the unique interpretability-specific losses. We show the application of our model on two publicly available datasets, Tadpole and the UK Biobank (UKBB). For Tadpole we choose the task of disease classification, and for UKBB, age, and sex prediction. The proposed model achieves an increase in an average accuracy of 3.2% for Tadpole and 1.6% for UKBB sex and 2% for the UKBB age prediction task compared to the state-of-the-art. Further, we show exhaustive validation and clinical interpretation of our results.

Graph Neural Networks usually rely on the assumption that the graph topology is available to the network as well as optimal for the downstream task. Latent graph inference allows models to dynamically learn the intrinsic graph structure of problems where the connectivity patterns of data may not be directly accessible. In this work, we generalize the discrete Differentiable Graph Module (dDGM) for latent graph learning. The original dDGM architecture used the Euclidean plane to encode latent features based on which the latent graphs were generated. By incorporating Riemannian geometry into the model and generating more complex embedding spaces, we can improve the performance of the latent graph inference system. In particular, we propose a computationally tractable approach to produce product manifolds of constant curvature model spaces that can encode latent features of varying structure. The latent representations mapped onto the inferred product manifold are used to compute richer similarity measures that are leveraged by the latent graph learning model to obtain optimized latent graphs. Moreover, the curvature of the product manifold is learned during training alongside the rest of the network parameters and based on the downstream task, rather than it being a static embedding space. Our novel approach is tested on a wide range of datasets, and outperforms the original dDGM model.

Pre-training has shown success in different areas of machine learning, such as Computer Vision (CV), Natural Language Processing (NLP) and medical imaging. However, it has not been fully explored for clinical data analysis. Even though an immense amount of Electronic Health Record (EHR) data is recorded, data and labels can be scarce if the data is collected in small hospitals or deals with rare diseases. In such scenarios, pre-training on a larger set of EHR data could improve the model performance. In this paper, we apply unsupervised pre-training to heterogeneous, multi-modal EHR data for patient outcome prediction. To model this data, we leverage graph deep learning over population graphs. We first design a network architecture based on graph transformer designed to handle various input feature types occurring in EHR data, like continuous, discrete, and time-series features, allowing better multi-modal data fusion. Further, we design pre-training methods based on masked imputation to pre-train our network before fine-tuning on different end tasks. Pre-training is done in a fully unsupervised fashion, which lays the groundwork for pre-training on large public datasets with different tasks and similar modalities in the future. We test our method on two medical datasets of patient records, TADPOLE and MIMIC-III, including imaging and non-imaging features and different prediction tasks. We find that our proposed graph based pre-training method helps in modeling the data at a population level and further improves performance on the fine tuning tasks in terms of AUC on average by 4.15% for MIMIC and 7.64% for TADPOLE.

Deep learning has revolutionized neuroimaging studies of psychiatric and neurological disorders. The difference between brain age and chronological age is a useful biomarker for identifying neurological disorders. Furthermore, delineating both age and gender is important for the study of illnesses exhibiting the phenotypic difference in these. In this paper, we focus on the prediction of age and gender from brain connectomes data which is a step further to full automation of disease prediction. We model the connectomes as brain graphs. Data is collected as functional MRI (fMRI) signals and the graphs are created by binarizing the correlation among the fMRI signals at the brain parcels considered as nodes. Such a graph represents the neurobiological functional connectivity. We further differentiate between static and dynamic connectivity. The former is constructed with the correlation of the overall signal at the nodes, while the latter is modeled as a sequence of brain graphs constructed over sequential time periods. Our hypothesis is that leveraging information from both the static and dynamic functional connectivity is beneficial to the task at hand. Our main contribution lies in our proposed novel input data representation and proposed recurrent graph-based deep learning model setting together. The proposed Dynamic Graph-based Gated Recurrent Unit (DG-GRU) comprises a mechanism to process both types of connectivities. In addition, it can be easily incorporated into any deep neural model. We show a thorough analysis of the model on two publicly available datasets HCP and ABIDE for two tasks to show the superiority of the model.

Large-scale population-based studies in medicine are a key resource towards better diagnosis, monitoring, and treatment of diseases. They also serve as enablers of clinical decision support systems, in particular computer-aided diagnosis (CADx) using machine learning (ML). Numerous ML approaches for CADx have been proposed in literature. However, these approaches assume feature-complete data, which is often not the case in clinical data. To account for missing data, incomplete data samples are either removed or imputed, which could lead to data bias and may negatively affect classification performance. As a solution, we propose an end-to-end learning of imputation and disease prediction of incomplete medical datasets via Multi-graph Geometric Matrix Completion (MGMC). MGMC uses multiple recurrent graph convolutional networks, where each graph represents an independent population model based on a key clinical meta-feature like age, sex, or cognitive function. Graph signal aggregation from local patient neighborhoods, combined with multi-graph signal fusion via self-attention, has a regularizing effect on both matrix reconstruction and classification performance. Our proposed approach is able to impute class relevant features as well as perform accurate and robust classification on two publicly available medical datasets. We empirically show the superiority of our proposed approach in terms of classification and imputation performance when compared with state-of-the-art approaches. MGMC enables disease prediction in multimodal and incomplete medical datasets. These findings could serve as baseline for future CADx approaches which utilize incomplete datasets.

The increased amount of multi-modal medical data has opened the opportunities to simultaneously process various modalities such as imaging and non-imaging data to gain a comprehensive insight into the disease prediction domain. Recent studies using Graph Convolutional Networks (GCNs) provide novel semi-supervised approaches for integrating heterogeneous modalities while investigating the patients' associations for disease prediction. However, when the meta-data used for graph construction is not available at inference time (e.g., coming from a distinct population), the conventional methods exhibit poor performance. To address this issue, we propose a novel semi-supervised approach named GKD based on knowledge distillation. We train a teacher component that employs the label-propagation algorithm besides a deep neural network to benefit from the graph and non-graph modalities only in the training phase. The teacher component embeds all the available information into the soft pseudo-labels. The soft pseudo-labels are then used to train a deep student network for disease prediction of unseen test data for which the graph modality is unavailable. We perform our experiments on two public datasets for diagnosing Autism spectrum disorder, and Alzheimer's disease, along with a thorough analysis on synthetic multi-modal datasets. According to these experiments, GKD outperforms the previous graph-based deep learning methods in terms of accuracy, AUC, and Macro F1.

Disease prediction is a well-known classification problem in medical applications. Graph Convolutional Networks (GCNs) provide a powerful tool for analyzing the patients' features relative to each other. This can be achieved by modeling the problem as a graph node classification task, where each node is a patient. Due to the nature of such medical datasets, class imbalance is a prevalent issue in the field of disease prediction, where the distribution of classes is skewed. When the class imbalance is present in the data, the existing graph-based classifiers tend to be biased towards the major class(es) and neglect the samples in the minor class(es). On the other hand, the correct diagnosis of the rare positive cases (true-positives) among all the patients is vital in a healthcare system. In conventional methods, such imbalance is tackled by assigning appropriate weights to classes in the loss function which is still dependent on the relative values of weights, sensitive to outliers, and in some cases biased towards the minor class(es). In this paper, we propose a Re-weighted Adversarial Graph Convolutional Network (RA-GCN) to prevent the graph-based classifier from emphasizing the samples of any particular class. This is accomplished by associating a graph-based neural network to each class, which is responsible for weighting the class samples and changing the importance of each sample for the classifier. Therefore, the classifier adjusts itself and determines the boundary between classes with more attention to the important samples. The parameters of the classifier and weighting networks are trained by an adversarial approach. We show experiments on synthetic and three publicly available medical datasets. Our results demonstrate the superiority of RA-GCN compared to recent methods in identifying the patient's status on all three datasets. The detailed analysis of our method is provided as quantitative and qualitative experiments on synthetic datasets.

Recently, Graph Convolutional Networks (GCNs) has proven to be a powerful machine learning tool for Computer Aided Diagnosis (CADx) and disease prediction. A key component in these models is to build a population graph, where the graph adjacency matrix represents pair-wise patient similarities. Until now, the similarity metrics have been defined manually, usually based on meta-features like demographics or clinical scores. The definition of the metric, however, needs careful tuning, as GCNs are very sensitive to the graph structure. In this paper, we demonstrate for the first time in the CADx domain that it is possible to learn a single, optimal graph towards the GCN's downstream task of disease classification. To this end, we propose a novel, end-to-end trainable graph learning architecture for dynamic and localized graph pruning. Unlike commonly employed spectral GCN approaches, our GCN is spatial and inductive, and can thus infer previously unseen patients as well. We demonstrate significant classification improvements with our learned graph on two CADx problems in medicine. We further explain and visualize this result using an artificial dataset, underlining the importance of graph learning for more accurate and robust inference with GCNs in medical applications.

Demonstrate the feasibility of a fully automatic computer-aided diagnosis (CAD) tool, based on deep learning, that localizes and classifies proximal femur fractures on X-ray images according to the AO classification. The proposed framework aims to improve patient treatment planning and provide support for the training of trauma surgeon residents. A database of 1347 clinical radiographic studies was collected. Radiologists and trauma surgeons annotated all fractures with bounding boxes and provided a classification according to the AO standard. In all experiments, the dataset was split patient-wise in three with the ratio 70%:10%:20% to build the training, validation and test sets, respectively. ResNet-50 and AlexNet architectures were implemented as deep learning classification and localization models, respectively. Accuracy, precision, recall and $$F_1$$-score were reported as classification metrics. Retrieval of similar cases was evaluated in terms of precision and recall. The proposed CAD tool for the classification of radiographs into types “A,” “B” and “not-fractured” reaches a $$F_1$$-score of 87% and AUC of 0.95. When classifying fractures versus not-fractured cases it improves up to 94% and 0.98. Prior localization of the fracture results in an improvement with respect to full-image classification. In total, 100% of the predicted centers of the region of interest are contained in the manually provided bounding boxes. The system retrieves on average 9 relevant images (from the same class) out of 10 cases. Our CAD scheme localizes, detects and further classifies proximal femur fractures achieving results comparable to expert-level and state-of-the-art performance. Our auxiliary localization model was highly accurate predicting the region of interest in the radiograph. We further investigated several strategies of verification for its adoption into the daily clinical routine. A sensitivity analysis of the size of the ROI and image retrieval as a clinical use case were presented.

Fractures of the proximal femur represent a critical entity in the western world, particularly with the growing elderly population. Such fractures result in high morbidity and mortality, reflecting a significant health and economic impact on our society. Different treatment strategies are recommended for different fracture types, with surgical treatment still being the gold standard in most of the cases. The success of the treatment and prognosis after surgery strongly depends on an accurate classification of the fracture among standard types, such as those defined by the AO system. However, the classification of fracture types based on x-ray images is difficult as confirmed by low intra- and inter-expert agreement rates of our in-house study and also in the previous literature. The presented work proposes a fully automatic computer-aided diagnosis (CAD) tool, based on current deep learning techniques, able to identify, localize and finally classify proximal femur fractures on x-rays images according to the AO classification. Results of our experimental evaluation show that the performance achieved by the proposed CAD tool is comparable to the average expert for the classification of x-ray images into types ''A'', ''B'' and ''normal'' (precision of 89%), while the performance is even superior when classifying fractures versus ''normal'' cases (precision of 94%). In addition, the integration of the proposed CAD tool into daily clinical routine is extensively discussed, towards improving the interface between humans and AI-powered machines in supporting medical decisions.