In recent years, data-driven machine learning (ML) methods have revolutionized the computer vision community by providing novel efficient solutions to many unsolved (medical) image analysis problems. However, due to the increasing privacy concerns and data fragmentation on many different sites, existing medical data are not fully utilized, thus limiting the potential of ML. Federated learning (FL) enables multiple parties to collaboratively train a ML model without exchanging local data. However, data heterogeneity (non-IID) among the distributed clients is yet a challenge. To this end, we propose a novel federated method, denoted Federated Disentanglement (FedDis), to disentangle the parameter space into shape and appearance, and only share the shape parameter with the clients. FedDis is based on the assumption that the anatomical structure in brain MRI images is similar across multiple institutions, and sharing the shape knowledge would be beneficial in anomaly detection. In this paper, we leverage healthy brain scans of 623 subjects from multiple sites with real data (OASIS, ADNI) in a privacy-preserving fashion to learn a model of normal anatomy, that allows to segment abnormal structures. We demonstrate a superior performance of FedDis on real pathological databases containing 109 subjects; two publicly available MS Lesions (MSLUB, MSISBI), and an in-house database with MS and Glioblastoma (MSI and GBI). FedDis achieved an average dice performance of 0.38, outperforming the state-of-the-art (SOTA) auto-encoder by 42% and the SOTA federated method by 11%. Further, we illustrate that FedDis learns a shape embedding that is orthogonal to the appearance and consistent under different intensity augmentations.
Standard machine learning approaches require centralizing the users' data in one computer or a shared database, which raises data privacy and confidentiality concerns. Therefore, limiting central access is important, especially in healthcare settings, where data regulations are strict. A potential approach to tackling this is Federated Learning (FL), which enables multiple parties to collaboratively learn a shared prediction model by using parameters of locally trained models while keeping raw training data locally. In the context of AI-assisted pain-monitoring, we wish to enable confidentiality-preserving and unobtrusive pain estimation for long-term pain-monitoring and reduce the burden on the nursing staff who perform frequent routine check-ups. To this end, we propose a novel Personalized Federated Deep Learning (PFDL) approach for pain estimation from face images. PFDL performs collaborative training of a deep model, implemented using a lightweight CNN architecture, across different clients (i.e., subjects) without sharing their face images. Instead of sharing all parameters of the model, as in standard FL, PFDL retains the last layer locally (used to personalize the pain estimates). This (i) adds another layer of data confidentiality, making it difficult for an adversary to infer pain levels of the target subject, while (ii) personalizing the pain estimation to each subject through local parameter tuning. We show using a publicly available dataset of face videos of pain (UNBC-McMaster Shoulder Pain Database), that PFDL performs comparably or better than the standard centralized and FL algorithms, while further enhancing data privacy. This, has the potential to improve traditional pain monitoring by making it more secure, computationally efficient, and scalable to a large number of individuals (e.g., for in-home pain monitoring), providing timely and unobtrusive pain measurement.
Purpose: To introduce a novel deep learning based approach for fast and high-quality dynamic multi-coil MR reconstruction by learning a complementary time-frequency domain network that exploits spatio-temporal correlations simultaneously from complementary domains. Theory and Methods: Dynamic parallel MR image reconstruction is formulated as a multi-variable minimisation problem, where the data is regularised in combined temporal Fourier and spatial (x-f) domain as well as in spatio-temporal image (x-t) domain. An iterative algorithm based on variable splitting technique is derived, which alternates among signal de-aliasing steps in x-f and x-t spaces, a closed-form point-wise data consistency step and a weighted coupling step. The iterative model is embedded into a deep recurrent neural network which learns to recover the image via exploiting spatio-temporal redundancies in complementary domains. Results: Experiments were performed on two datasets of highly undersampled multi-coil short-axis cardiac cine MRI scans. Results demonstrate that our proposed method outperforms the current state-of-the-art approaches both quantitatively and qualitatively. The proposed model can also generalise well to data acquired from a different scanner and data with pathologies that were not seen in the training set. Conclusion: The work shows the benefit of reconstructing dynamic parallel MRI in complementary time-frequency domains with deep neural networks. The method can effectively and robustly reconstruct high-quality images from highly undersampled dynamic multi-coil data ($16 \times$ and $24 \times$ yielding 15s and 10s scan times respectively) with fast reconstruction speed (2.8s). This could potentially facilitate achieving fast single-breath-hold clinical 2D cardiac cine imaging.
The utilisation of artificial intelligence in medicine and healthcare has led to successful clinical applications in several domains. The conflict between data usage and privacy protection requirements in such systems must be resolved for optimal results as well as ethical and legal compliance. This calls for innovative solutions such as privacy-preserving machine learning (PPML). We present PriMIA (Privacy-preserving Medical Image Analysis), a software framework designed for PPML in medical imaging. In a real-life case study we demonstrate significantly better classification performance of a securely aggregated federated learning model compared to human experts on unseen datasets. Furthermore, we show an inference-as-a-service scenario for end-to-end encrypted diagnosis, where neither the data nor the model are revealed. Lastly, we empirically evaluate the framework's security against a gradient-based model inversion attack and demonstrate that no usable information can be recovered from the model.
Despite current advances in deep learning, domain shift remains a common problem in medical imaging settings. Recent findings on natural images suggest that deep neural models can show a textural bias when carrying out image classification tasks, which goes against the common understanding of convolutional neural networks (CNNs) recognising objects through increasingly complex representations of shape. This study draws inspiration from recent findings on natural images and aims to investigate ways in which addressing the textural bias phenomenon could be used to bring up the robustness and transferability of deep segmentation models when applied to three-dimensional (3D) medical data. To achieve this, publicly available MRI scans from the Developing Human Connectome Project are used to investigate ways in which simulating textural noise can help train robust models in a complex segmentation task. Our findings illustrate how applying specific types of textural filters prior to training the models can increase their ability to segment scans corrupted by previously unseen noise.
Federated Learning harnesses data from multiple sources to build a single model. While the initial model might belong solely to the actor bringing it to the network for training, determining the ownership of the trained model resulting from Federated Learning remains an open question. In this paper we explore how Blockchains (in particular Ethereum) can be used to determine the evolving ownership of a model trained with Federated Learning. Firstly, we use the step-by-step evaluation metric to assess the relative contributivities of participants in a Federated Learning process. Next, we introduce 2CP, a framework comprising two novel protocols for Blockchained Federated Learning, which both reward contributors with shares in the final model based on their relative contributivity. The Crowdsource Protocol allows an actor to bring a model forward for training, and use their own data to evaluate the contributions made to it. Potential trainers are guaranteed a fair share of the resulting model, even in a trustless setting. The Consortium Protocol gives trainers the same guarantee even when no party owns the initial model and no evaluator is available. We conduct experiments with the MNIST dataset that reveal sound contributivity scores resulting from both Protocols by rewarding larger datasets with greater shares in the model. Our experiments also showed the necessity to pair 2CP with a robust model aggregation mechanism to discard low quality inputs coming from model poisoning attacks.
This work provides a comprehensive review of existing frameworks based on secure computing techniques in the context of private image classification. The in-depth analysis of these approaches is followed by careful examination of their performance costs, in particular runtime and communication overhead. To further illustrate the practical considerations when using different privacy-preserving technologies, experiments were conducted using four state-of-the-art libraries implementing secure computing at the heart of the data science stack: PySyft and CrypTen supporting private inference via Secure Multi-Party Computation, TF-Trusted utilising Trusted Execution Environments and HE- Transformer relying on Homomorphic encryption. Our work aims to evaluate the suitability of these frameworks from a usability, runtime requirements and accuracy point of view. In order to better understand the gap between state-of-the-art protocols and what is currently available in practice for a data scientist, we designed three neural network architecture to obtain secure predictions via each of the four aforementioned frameworks. Two networks were evaluated on the MNIST dataset and one on the Malaria Cell image dataset. We observed satisfying performances for TF-Trusted and CrypTen and noted that all frameworks perfectly preserved the accuracy of the corresponding plaintext model.
Deep neural networks exhibit limited generalizability across images with different entangled domain features and categorical features. Learning generalizable features that can form universal categorical decision boundaries across domains is an interesting and difficult challenge. This problem occurs frequently in medical imaging applications when attempts are made to deploy and improve deep learning models across different image acquisition devices, across acquisition parameters or if some classes are unavailable in new training databases. To address this problem, we propose Mutual Information-based Disentangled Neural Networks (MIDNet), which extract generalizable categorical features to transfer knowledge to unseen categories in a target domain. The proposed MIDNet adopts a semi-supervised learning paradigm to alleviate the dependency on labeled data. This is important for real-world applications where data annotation is time-consuming, costly and requires training and expertise. We extensively evaluate the proposed method on fetal ultrasound datasets for two different image classification tasks where domain features are respectively defined by shadow artifacts and image acquisition devices. Experimental results show that the proposed method outperforms the state-of-the-art on the classification of unseen categories in a target domain with sparsely labeled training data.
Brain age estimation from Magnetic Resonance Images (MRI) derives the difference between a subject's biological brain age and their chronological age. This is a potential biomarker for neurodegeneration, e.g. as part of Alzheimer's disease. Early detection of neurodegeneration manifesting as a higher brain age can potentially facilitate better medical care and planning for affected individuals. Many studies have been proposed for the prediction of chronological age from brain MRI using machine learning and specifically deep learning techniques. Contrary to most studies, which use the whole brain volume, in this study, we develop a new deep learning approach that uses 3D patches of the brain as well as convolutional neural networks (CNNs) to develop a localised brain age estimator. In this way, we can obtain a visualization of the regions that play the most important role for estimating brain age, leading to more anatomically driven and interpretable results, and thus confirming relevant literature which suggests that the ventricles and the hippocampus are the areas that are most informative. In addition, we leverage this knowledge in order to improve the overall performance on the task of age estimation by combining the results of different patches using an ensemble method, such as averaging or linear regression. The network is trained on the UK Biobank dataset and the method achieves state-of-the-art results with a Mean Absolute Error of 2.46 years for purely regional estimates, and 2.13 years for an ensemble of patches before bias correction, while 1.96 years after bias correction.
The estimation of glaucoma progression is a challenging task as the rate of disease progression varies among individuals in addition to other factors such as measurement variability and the lack of standardization in defining progression. Structural tests, such as thickness measurements of the retinal nerve fiber layer or the macula with optical coherence tomography (OCT), are able to detect anatomical changes in glaucomatous eyes. Such changes may be observed before any functional damage. In this work, we built a generative deep learning model using the conditional GAN architecture to predict glaucoma progression over time. The patient's OCT scan is predicted from three or two prior measurements. The predicted images demonstrate high similarity with the ground truth images. In addition, our results suggest that OCT scans obtained from only two prior visits may actually be sufficient to predict the next OCT scan of the patient after six months.