Counterfactual image generation is pivotal for understanding the causal relations of variables, with applications in interpretability and generation of unbiased synthetic data. However, evaluating image generation is a long-standing challenge in itself. The need to evaluate counterfactual generation compounds on this challenge, precisely because counterfactuals, by definition, are hypothetical scenarios without observable ground truths. In this paper, we present a novel comprehensive framework aimed at benchmarking counterfactual image generation methods. We incorporate metrics that focus on evaluating diverse aspects of counterfactuals, such as composition, effectiveness, minimality of interventions, and image realism. We assess the performance of three distinct conditional image generation model types, based on the Structural Causal Model paradigm. Our work is accompanied by a user-friendly Python package which allows to further evaluate and benchmark existing and future counterfactual image generation methods. Our framework is extendable to additional SCM and other causal methods, generative models, and datasets.
Deep learning (DL) has substantially enhanced healthcare research by addressing various natural language processing (NLP) tasks. Yet, the increasing complexity of DL-based NLP methods necessitates transparent model interpretability, or at least explainability, for reliable decision-making. This work presents a thorough scoping review on explainable and interpretable DL in healthcare NLP. The term "XIAI" (eXplainable and Interpretable Artificial Intelligence) was introduced to distinguish XAI from IAI. Methods were further categorized based on their functionality (model-, input-, output-based) and scope (local, global). Our analysis shows that attention mechanisms were the most dominant emerging IAI. Moreover, IAI is increasingly used against XAI. The major challenges identified are that most XIAI do not explore "global" modeling processes, the lack of best practices, and the unmet need for systematic evaluation and benchmarks. Important opportunities were raised such as using "attention" to enhance multi-modal XIAI for personalized medicine and combine DL with causal reasoning. Our discussion encourages the integration of XIAI in LLMs and domain-specific smaller models. Our review can stimulate further research and benchmarks toward improving inherent IAI and engaging complex NLP in healthcare.
Clinical decision making from magnetic resonance imaging (MRI) combines complementary information from multiple MRI sequences (defined as 'modalities'). MRI image registration aims to geometrically 'pair' diagnoses from different modalities, time points and slices. Both intra- and inter-modality MRI registration are essential components in clinical MRI settings. Further, an MRI image processing pipeline that can address both afine and non-rigid registration is critical, as both types of deformations may be occuring in real MRI data scenarios. Unlike image classification, explainability is not commonly addressed in image registration deep learning (DL) methods, as it is challenging to interpet model-data behaviours against transformation fields. To properly address this, we incorporate Grad-CAM-based explainability frameworks in each major component of our unsupervised multi-modal and multi-organ image registration DL methodology. We previously demonstrated that we were able to reach superior performance (against the current standard Syn method). In this work, we show that our DL model becomes fully explainable, setting the framework to generalise our approach on further medical imaging data.
Medical imaging is a key component in clinical diagnosis, treatment planning and clinical trial design, accounting for almost 90% of all healthcare data. CNNs achieved performance gains in medical image analysis (MIA) over the last years. CNNs can efficiently model local pixel interactions and be trained on small-scale MI data. The main disadvantage of typical CNN models is that they ignore global pixel relationships within images, which limits their generalisation ability to understand out-of-distribution data with different 'global' information. The recent progress of Artificial Intelligence gave rise to Transformers, which can learn global relationships from data. However, full Transformer models need to be trained on large-scale data and involve tremendous computational complexity. Attention and Transformer compartments (Transf/Attention) which can well maintain properties for modelling global relationships, have been proposed as lighter alternatives of full Transformers. Recently, there is an increasing trend to co-pollinate complementary local-global properties from CNN and Transf/Attention architectures, which led to a new era of hybrid models. The past years have witnessed substantial growth in hybrid CNN-Transf/Attention models across diverse MIA problems. In this systematic review, we survey existing hybrid CNN-Transf/Attention models, review and unravel key architectural designs, analyse breakthroughs, and evaluate current and future opportunities as well as challenges. We also introduced a comprehensive analysis framework on generalisation opportunities of scientific and clinical impact, based on which new data-driven domain generalisation and adaptation methods can be stimulated.
As a pragmatic data augmentation tool, data synthesis has generally returned dividends in performance for deep learning based medical image analysis. However, generating corresponding segmentation masks for synthetic medical images is laborious and subjective. To obtain paired synthetic medical images and segmentations, conditional generative models that use segmentation masks as synthesis conditions were proposed. However, these segmentation mask-conditioned generative models still relied on large, varied, and labeled training datasets, and they could only provide limited constraints on human anatomical structures, leading to unrealistic image features. Moreover, the invariant pixel-level conditions could reduce the variety of synthetic lesions and thus reduce the efficacy of data augmentation. To address these issues, in this work, we propose a novel strategy for medical image synthesis, namely Unsupervised Mask (UM)-guided synthesis, to obtain both synthetic images and segmentations using limited manual segmentation labels. We first develop a superpixel based algorithm to generate unsupervised structural guidance and then design a conditional generative model to synthesize images and annotations simultaneously from those unsupervised masks in a semi-supervised multi-task setting. In addition, we devise a multi-scale multi-task Fr\'echet Inception Distance (MM-FID) and multi-scale multi-task standard deviation (MM-STD) to harness both fidelity and variety evaluations of synthetic CT images. With multiple analyses on different scales, we could produce stable image quality measurements with high reproducibility. Compared with the segmentation mask guided synthesis, our UM-guided synthesis provided high-quality synthetic images with significantly higher fidelity, variety, and utility ($p<0.05$ by Wilcoxon Signed Ranked test).
Magnetic Resonance Imaging (MRI) typically recruits multiple sequences (defined here as "modalities"). As each modality is designed to offer different anatomical and functional clinical information, there are evident disparities in the imaging content across modalities. Inter- and intra-modality affine and non-rigid image registration is an essential medical image analysis process in clinical imaging, as for example before imaging biomarkers need to be derived and clinically evaluated across different MRI modalities, time phases and slices. Although commonly needed in real clinical scenarios, affine and non-rigid image registration is not extensively investigated using a single unsupervised model architecture. In our work, we present an un-supervised deep learning registration methodology which can accurately model affine and non-rigid trans-formations, simultaneously. Moreover, inverse-consistency is a fundamental inter-modality registration property that is not considered in deep learning registration algorithms. To address inverse-consistency, our methodology performs bi-directional cross-modality image synthesis to learn modality-invariant latent rep-resentations, while involves two factorised transformation networks and an inverse-consistency loss to learn topology-preserving anatomical transformations. Overall, our model (named "FIRE") shows improved performances against the reference standard baseline method on multi-modality brain 2D and 3D MRI and intra-modality cardiac 4D MRI data experiments.
Automated pathology segmentation remains a valuable diagnostic tool in clinical practice. However, collecting training data is challenging. Semi-supervised approaches by combining labelled and unlabelled data can offer a solution to data scarcity. An approach to semi-supervised learning relies on reconstruction objectives (as self-supervision objectives) that learns in a joint fashion suitable representations for the task. Here, we propose Anatomy-Pathology Disentanglement Network (APD-Net), a pathology segmentation model that attempts to learn jointly for the first time: disentanglement of anatomy, modality, and pathology. The model is trained in a semi-supervised fashion with new reconstruction losses directly aiming to improve pathology segmentation with limited annotations. In addition, a joint optimization strategy is proposed to fully take advantage of the available annotations. We evaluate our methods with two private cardiac infarction segmentation datasets with LGE-MRI scans. APD-Net can perform pathology segmentation with few annotations, maintain performance with different amounts of supervision, and outperform related deep learning methods.
Magnetic resonance (MR) protocols rely on several sequences to properly assess pathology and organ status. Yet, despite advances in image analysis we tend to treat each sequence, here termed modality, in isolation. Taking advantage of the information shared between modalities (largely an organ's anatomy) is beneficial for multi-modality multi-input processing and learning. However, we must overcome inherent anatomical misregistrations and disparities in signal intensity across the modalities to claim this benefit. We present a method that offers improved segmentation accuracy of the modality of interest (over a single input model), by learning to leverage information present in other modalities, enabling semi-supervised and zero shot learning. Core to our method is learning a disentangled decomposition into anatomical and imaging factors. Shared anatomical factors from the different inputs are jointly processed and fused to extract more accurate segmentation masks. Image misregistrations are corrected with a Spatial Transformer Network, that non-linearly aligns the anatomical factors. The imaging factor captures signal intensity characteristics across different modality data, and is used for image reconstruction, enabling semi-supervised learning. Temporal and slice pairing between inputs are learned dynamically. We demonstrate applications in Late Gadolinium Enhanced (LGE) and Blood Oxygenation Level Dependent (BOLD) cardiac segmentation, as well as in T2 abdominal segmentation.
Inter-modality image registration is an critical preprocessing step for many applications within the routine clinical pathway. This paper presents an unsupervised deep inter-modality registration network that can learn the optimal affine and non-rigid transformations simultaneously. Inverse-consistency is an important property commonly ignored in recent deep learning based inter-modality registration algorithms. We address this issue through the proposed multi-task architecture and the new comprehensive transformation network. Specifically, the proposed model learns a modality-independent latent representation to perform cycle-consistent cross-modality synthesis, and use an inverse-consistent loss to learn a pair of transformations to align the synthesized image with the target. We name this proposed framework as FIRE due to the shape of its structure. Our method shows comparable and better performances with the popular baseline method in experiments on multi-sequence brain MR data and intra-modality 4D cardiac Cine-MR data.
Typically, a medical image offers spatial information on the anatomy (and pathology) modulated by imaging specific characteristics. Many imaging modalities including Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) can be interpreted in this way. We can venture further and consider that a medical image naturally factors into some spatial factors depicting anatomy and factors that denote the imaging characteristics. Here, we explicitly learn this decomposed (factorised) representation of imaging data, focusing in particular on cardiac images. We propose Spatial Decomposition Network (SDNet), which factorises 2D medical images into spatial anatomical factors and non-spatial imaging factors. We demonstrate that this high-level representation is ideally suited for several medical image analysis tasks, such as semi-supervised segmentation, multi-task segmentation and regression, and image-to-image synthesis. Specifically, we show that our model can match the performance of fully supervised segmentation models, using only a fraction of the labelled images. Critically, we show that our factorised representation also benefits from supervision obtained either when we use auxiliary tasks to train the model in a multi-task setting (e.g. regressing to known cardiac indices), or when aggregating multimodal data from different sources (e.g. pooling together MRI and CT data). To explore the properties of the learned factorisation, we perform latent-space arithmetic and show that we can synthesise CT from MR and vice versa, by swapping the modality factors. We also demonstrate that the factor holding image specific information can be used to predict the input modality with high accuracy.