on behalf of the PINNACLE consortium
Abstract:Clinicians spend a significant amount of time reviewing medical images and transcribing their findings regarding patient diagnosis, referral and treatment in text form. Vision-language models (VLMs), which automatically interpret images and summarize their findings as text, have enormous potential to alleviate clinical workloads and increase patient access to high-quality medical care. While foundational models have stirred considerable interest in the medical community, it is unclear whether their general capabilities translate to real-world clinical utility. In this work, we show that foundation VLMs markedly underperform compared to practicing ophthalmologists on specialist tasks crucial to the care of patients with age-related macular degeneration (AMD). To address this, we initially identified the essential capabilities required for image-based clinical decision-making, and then developed a curriculum to selectively train VLMs in these skills. The resulting model, RetinaVLM, can be instructed to write reports that significantly outperform those written by leading foundation medical VLMs in disease staging (F1 score of 0.63 vs. 0.11) and patient referral (0.67 vs. 0.39), and approaches the diagnostic performance of junior ophthalmologists (who achieve 0.77 and 0.78 on the respective tasks). Furthermore, in a reader study involving two senior ophthalmologists with up to 32 years of experience, RetinaVLM's reports were found to be similarly correct (78.6% vs. 82.1%) and complete (both 78.6%) as reports written by junior ophthalmologists with up to 10 years of experience. These results demonstrate that our curriculum-based approach provides a blueprint for specializing generalist foundation medical VLMs to handle real-world clinical tasks.
Abstract:Machine unlearning is the process of removing the impact of a particular set of training samples from a pretrained model. It aims to fulfill the "right to be forgotten", which grants the individuals such as patients the right to reconsider their contribution in models including medical imaging models. In this study, we evaluate the effectiveness (performance) and computational efficiency of different unlearning algorithms in medical imaging domain. Our evaluations demonstrate that the considered unlearning algorithms perform well on the retain set (samples whose influence on the model is allowed to be retained) and forget set (samples whose contribution to the model should be eliminated), and show no bias against male or female samples. They, however, adversely impact the generalization of the model, especially for larger forget set sizes. Moreover, they might be biased against easy or hard samples, and need additional computational overhead for hyper-parameter tuning. In conclusion, machine unlearning seems promising for medical imaging, but the existing unlearning algorithms still needs further improvements to become more practical for medical applications.
Abstract:With the increasing incidence of neurodegenerative diseases such as Alzheimer's Disease (AD), there is a need for further research that enhances detection and monitoring of the diseases. We present MORPHADE (Morphological Autoencoders for Alzheimer's Disease Detection), a novel unsupervised learning approach which uses deformations to allow the analysis of 3D T1-weighted brain images. To the best of our knowledge, this is the first use of deformations with deep unsupervised learning to not only detect, but also localize and assess the severity of structural changes in the brain due to AD. We obtain markedly higher anomaly scores in clinically important areas of the brain in subjects with AD compared to healthy controls, showcasing that our method is able to effectively locate AD-related atrophy. We additionally observe a visual correlation between the severity of atrophy highlighted in our anomaly maps and medial temporal lobe atrophy scores evaluated by a clinical expert. Finally, our method achieves an AUROC of 0.80 in detecting AD, out-performing several supervised and unsupervised baselines. We believe our framework shows promise as a tool towards improved understanding, monitoring and detection of AD. To support further research and application, we have made our code publicly available at github.com/ci-ber/MORPHADE.
Abstract:Cardiac Cine Magnetic Resonance Imaging (MRI) provides an accurate assessment of heart morphology and function in clinical practice. However, MRI requires long acquisition times, with recent deep learning-based methods showing great promise to accelerate imaging and enhance reconstruction quality. Existing networks exhibit some common limitations that constrain further acceleration possibilities, including single-domain learning, reliance on a single regularization term, and equal feature contribution. To address these limitations, we propose to embed information from multiple domains, including low-rank, image, and k-space, in a novel deep learning network for MRI reconstruction, which we denote as A-LIKNet. A-LIKNet adopts a parallel-branch structure, enabling independent learning in the k-space and image domain. Coupled information sharing layers realize the information exchange between domains. Furthermore, we introduce attention mechanisms into the network to assign greater weights to more critical coils or important temporal frames. Training and testing were conducted on an in-house dataset, including 91 cardiovascular patients and 38 healthy subjects scanned with 2D cardiac Cine using retrospective undersampling. Additionally, we evaluated A-LIKNet on the real-time 8x prospectively undersampled data from the OCMR dataset. The results demonstrate that our proposed A-LIKNet outperforms existing methods and provides high-quality reconstructions. The network can effectively reconstruct highly retrospectively undersampled dynamic MR images up to 24x accelerations, indicating its potential for single breath-hold imaging.
Abstract:Segmentation is often the first step in many medical image analyses workflows. Deep learning approaches, while giving state-of-the-art accuracies, are data intensive and do not scale well to low data regimes. We introduce Deep Conditional Shape Models 2.0, which uses an edge detector, along with an implicit shape function conditioned on edge maps, to leverage cross-modality shape information. The shape function is trained exclusively on a source domain (contrasted CT) and applied to the target domain of interest (3D echocardiography). We demonstrate data efficiency in the target domain by varying the amounts of training data used in the edge detection stage. We observe that DCSM 2.0 outperforms the baseline at all data levels in terms of Hausdorff distances, and while using 50% or less of the training data in terms of average mesh distance, and at 10% or less of the data with the dice coefficient. The method scales well to low data regimes, with gains of up to 5% in dice coefficient, 2.58 mm in average surface distance and 21.02 mm in Hausdorff distance when using just 2% (22 volumes) of the training data.
Abstract:Since the emergence of the Transformer architecture, language model development has increased, driven by their promising potential. However, releasing these models into production requires properly understanding their behavior, particularly in sensitive domains such as medicine. Despite this need, the medical literature still lacks technical assessments of pre-trained language models, which are especially valuable in resource-constrained settings in terms of computational power or limited budget. To address this gap, we provide a comprehensive survey of language models in the medical domain. In addition, we selected a subset of these models for thorough evaluation, focusing on classification and text generation tasks. Our subset encompasses 53 models, ranging from 110 million to 13 billion parameters, spanning the three families of Transformer-based models and from diverse knowledge domains. This study employs a series of approaches for text classification together with zero-shot prompting instead of model training or fine-tuning, which closely resembles the limited resource setting in which many users of language models find themselves. Encouragingly, our findings reveal remarkable performance across various tasks and datasets, underscoring the latent potential of certain models to contain medical knowledge, even without domain specialization. Consequently, our study advocates for further exploration of model applications in medical contexts, particularly in resource-constrained settings. The code is available on https://github.com/anpoc/Language-models-in-medicine.
Abstract:Existing learning-based cortical surface reconstruction approaches heavily rely on the supervision of pseudo ground truth (pGT) cortical surfaces for training. Such pGT surfaces are generated by traditional neuroimage processing pipelines, which are time consuming and difficult to generalize well to low-resolution brain MRI, e.g., from fetuses and neonates. In this work, we present CoSeg, a learning-based cortical surface reconstruction framework weakly supervised by brain segmentations without the need for pGT surfaces. CoSeg introduces temporal attention networks to learn time-varying velocity fields from brain MRI for diffeomorphic surface deformations, which fit an initial surface to target cortical surfaces within only 0.11 seconds for each brain hemisphere. A weakly supervised loss is designed to reconstruct pial surfaces by inflating the white surface along the normal direction towards the boundary of the cortical gray matter segmentation. This alleviates partial volume effects and encourages the pial surface to deform into deep and challenging cortical sulci. We evaluate CoSeg on 1,113 adult brain MRI at 1mm and 2mm resolution. CoSeg achieves superior geometric and morphological accuracy compared to existing learning-based approaches. We also verify that CoSeg can extract high-quality cortical surfaces from fetal brain MRI on which traditional pipelines fail to produce acceptable results.
Abstract:In differentially private (DP) machine learning, the privacy guarantees of DP mechanisms are often reported and compared on the basis of a single $(\varepsilon, \delta)$-pair. This practice overlooks that DP guarantees can vary substantially \emph{even between mechanisms sharing a given $(\varepsilon, \delta)$}, and potentially introduces privacy vulnerabilities which can remain undetected. This motivates the need for robust, rigorous methods for comparing DP guarantees in such cases. Here, we introduce the $\Delta$-divergence between mechanisms which quantifies the worst-case excess privacy vulnerability of choosing one mechanism over another in terms of $(\varepsilon, \delta)$, $f$-DP and in terms of a newly presented Bayesian interpretation. Moreover, as a generalisation of the Blackwell theorem, it is endowed with strong decision-theoretic foundations. Through application examples, we show that our techniques can facilitate informed decision-making and reveal gaps in the current understanding of privacy risks, as current practices in DP-SGD often result in choosing mechanisms with high excess privacy vulnerabilities.
Abstract:Field-of-view (FOV) recovery of truncated chest CT scans is crucial for accurate body composition analysis, which involves quantifying skeletal muscle and subcutaneous adipose tissue (SAT) on CT slices. This, in turn, enables disease prognostication. Here, we present a method for recovering truncated CT slices using generative image outpainting. We train a diffusion model and apply it to truncated CT slices generated by simulating a small FOV. Our model reliably recovers the truncated anatomy and outperforms the previous state-of-the-art despite being trained on 87% less data.
Abstract:Visual prompting infuses visual information into the input image to adapt models toward specific predictions and tasks. Recently, manually crafted markers such as red circles are shown to guide the model to attend to a target region on the image. However, these markers only work on models trained with data containing those markers. Moreover, finding these prompts requires guesswork or prior knowledge of the domain on which the model is trained. This work circumvents manual design constraints by proposing to learn the visual prompts for guiding the attention of vision transformers. The learned visual prompt, added to any input image would redirect the attention of the pre-trained vision transformer to its spatial location on the image. Specifically, the prompt is learned in a self-supervised manner without requiring annotations and without fine-tuning the vision transformer. Our experiments demonstrate the effectiveness of the proposed optimization-based visual prompting strategy across various pre-trained vision encoders.