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 (disentangled) 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 modality 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. Code will be made available at https://github. com/agis85/anatomy_modality_decomposition.

No abstract

Word Count: 1,561 (Including References) AbstractThe world is currently suffering

Elastin degradation is a key feature of emphysema and may have a role in the pathogenesis of atherosclerosis associated with chronic obstructive pulmonary disease (COPD). Circulating desmosine is a specific biomarker of elastin degradation. We investigated the association between plasma desmosine (pDES) and emphysema severity/progression, coronary artery calcium score (CACS) and mortality.pDES was measured in 1177 COPD patients and 110 healthy control subjects from two independent cohorts. Emphysema was assessed on chest computed tomography scans. Aortic arterial stiffness was measured as the aortic-femoral pulse wave velocity.pDES was elevated in patients with cardiovascular disease (p<0.005) and correlated with age (rho=0.39, p<0.0005), CACS (rho=0.19, p<0.0005) modified Medical Research Council dyspnoea score (rho=0.15, p<0.0005), 6-min walking distance (rho=-0.17, p<0.0005) and body mass index, airflow obstruction, dyspnoea, exercise capacity index (rho=0.10, p<0.01), but not with emphysema, emphysema progression or forced expiratory volume in 1 s decline. pDES predicted all-cause mortality independently of several confounding factors (p<0.005). In an independent cohort of 186 patients with COPD and 110 control subjects, pDES levels were higher in COPD patients with cardiovascular disease and correlated with arterial stiffness (p<0.05).In COPD, excess elastin degradation relates to cardiovascular comorbidities, atherosclerosis, arterial stiffness, systemic inflammation and mortality, but not to emphysema or emphysema progression. pDES is a good biomarker of cardiovascular risk and mortality in COPD.

Persistent ill health after acute COVID-19—referred to as long COVID, the post-acute COVID-19 syndrome, or the post-COVID-19 condition—has emerged as a major concern. We undertook an international consensus exercise to identify research priorities with the aim of understanding the long-term effects of acute COVID-19, with a focus on people with pre-existing airways disease and the occurrence of new-onset airways disease and associated symptoms. 202 international experts were invited to submit a minimum of three research ideas. After a two-phase internal review process, a final list of 98 research topics was scored by 48 experts. Patients with pre-existing or post-COVID-19 airways disease contributed to the exercise by weighting selected criteria. The highest-ranked research idea focused on investigation of the relationship between prognostic scores at hospital admission and morbidity at 3 months and 12 months after hospital discharge in patients with and without pre-existing airways disease. High priority was also assigned to comparisons of the prevalence and severity of post-COVID-19 fatigue, sarcopenia, anxiety, depression, and risk of future cardiovascular complications in patients with and without pre-existing airways disease. Our approach has enabled development of a set of priorities that could inform future research studies and funding decisions. This prioritisation process could also be adapted to other, non-respiratory aspects of long COVID.

No abstract

No abstract

The success and generalisation of deep learning algorithms heavily depend on learning good feature representations. In medical imaging this entails representing anatomical information, as well as properties related to the specific imaging setting. Anatomical information is required to perform further analysis, whereas imaging information is key to disentangle scanner variability and potential artefacts. The ability to factorise these would allow for training algorithms only on the relevant information according to the task. To date, such factorisation has not been attempted. In this paper, we propose a methodology of latent space factorisation relying on the cycle-consistency principle. As an example application, we consider cardiac MR segmentation, where we separate information related to the myocardium from other features related to imaging and surrounding substructures. We demonstrate the proposed method's utility in a semi-supervised setting: we use very few labelled images together with many unlabelled images to train a myocardium segmentation neural network. Specifically, we achieve comparable performance to fully supervised networks using a fraction of labelled images in experiments on ACDC and a dataset from Edinburgh Imaging Facility QMRI. Code will be made available at