Phenotypic Multiorgan Involvement of Subclinical Disease as Quantified by Magnetic Resonance Imaging in Subjects With Prediabetes, Diabetes, and Normal Glucose Tolerance:

Deep Learning-Based Automated Abdominal Organ Segmentation in the UK Biobank and German National Cohort Magnetic Resonance Imaging Studies

PURPOSE

The aims of this study were to train and evaluate deep learning models for automated segmentation of abdominal organs in whole-body magnetic resonance (MR) images from the UK Biobank (UKBB) and German National Cohort (GNC) MR imaging studies and to make these models available to the scientific community for analysis of these data sets.

METHODS

A total of 200 T1-weighted MR image data sets of healthy volunteers each from UKBB and GNC (400 data sets in total) were available in this study. Liver, spleen, left and right kidney, and pancreas were segmented manually on all 400 data sets, providing labeled ground truth data for training of a previously described U-Net-based deep learning framework for automated medical image segmentation (nnU-Net). The trained models were tested on all data sets using a 4-fold cross-validation scheme. Qualitative analysis of automated segmentation results was performed visually; performance metrics between automated and manual segmentation results were computed for quantitative analysis. In addition, interobserver segmentation variability between 2 human readers was assessed on a subset of the data.

RESULTS

Automated abdominal organ segmentation was performed with high qualitative and quantitative accuracy on UKBB and GNC data. In more than 90% of data sets, no or only minor visually detectable qualitative segmentation errors occurred. Mean Dice scores of automated segmentations compared with manual reference segmentations were well higher than 0.9 for the liver, spleen, and kidneys on UKBB and GNC data and around 0.82 and 0.89 for the pancreas on UKBB and GNC data, respectively. Mean average symmetric surface distance was between 0.3 and 1.5 mm for the liver, spleen, and kidneys and between 2 and 2.2 mm for pancreas segmentation. The quantitative accuracy of automated segmentation was comparable with the agreement between 2 human readers for all organs on UKBB and GNC data.

CONCLUSION

Automated segmentation of abdominal organs is possible with high qualitative and quantitative accuracy on whole-body MR imaging data acquired as part of UKBB and GNC. The results obtained and deep learning models trained in this study can be used as a foundation for automated analysis of thousands of MR data sets of UKBB and GNC and thus contribute to tackling topical and original scientific questions.

Brain deficits in prediabetic adults: A systematic review

Previous findings on the relationship between prediabetes (the precursor stage of type 2 diabetes) and brain health in humans are inconsistent. Thus, this systematic review of cross-sectional and longitudinal studies aimed to summarize what is currently known about brain deficits in prediabetic adults. Following the PRISMA reporting standards for systematic reviews, we conducted a comprehensive review of peer-reviewed journal articles published from 2009 to present, focusing on studies that assessed brain volume, structural connectivity, and cerebrovascular health in prediabetic adults and older adults (i.e., 18 years or older). We systematically searched PsychINFO, Scopus, Web of Science, Ovid MEDLINE, CINAHL, and EMbase databases. Quality assessment was based on the NIH Quality Assessment Tool for Observational and Cross-sectional Studies. In total, 19 studies were included in our review. Results from these studies show that prediabetes may be associated with deficits in brain structure and pathology, however, several studies also refute these findings. Moreover, we identified clear inconsistencies in study methodologies, including diabetes measures and classification, across studies that may account for these conflicting findings.

Prediabetes Is Independently Associated with Subclinical Carotid Atherosclerosis: An Observational Study in a Non-Urban Mediterranean Population

This was a prospective, observational study to compare the burden of subclinical atherosclerosis as measured by carotid ultrasonography in a cohort of subjects with prediabetes vs. subjects with normal glucose tolerance (NGT) from a non-urban Mediterranean population. Atherosclerosis was assessed through carotid intima-media thickness (c-IMT), the presence/absence of carotid plaques, and plaque number. Among 550 subjects included, 224 (40.7%) had prediabetes. The mean c-IMT and the prevalence of carotid plaque were significantly higher in the prediabetes group compared to the NGT group (0.72 vs. 0.67 mm, p < 0.001; and 37.9% vs. 19.6%; p < 0.001, respectively). Older age, male gender, and increased systolic blood pressure were positively correlated with c-IMT and were independent predictors of the presence of plaques. In contrast, prediabetes and low-density lipoprotein (LDL)-c were predictors of the presence of plaque (odds ratio [OR] = 1.64; 95% confidence interval [CI] = 1.05-2.57; p = 0.03 and OR = 1.01; 95% CI = 1.00-1.02; p = 0.006, respectively) together with tobacco exposure and the leukocyte count (OR = 1.77; 95% CI = 1.08-2.89; p = 0.023 and OR = 1.20; 95% CI = 1.05-1.38; p = 0.008, respectively). In a non-urban Mediterranean population, prediabetes was associated with established subclinical carotid atherosclerosis. These findings could have implications for the prevention and treatment of CV risk in these subjects before the first symptoms of cardiovascular disease appear.

Renal and renal sinus fat volumes as quantified by magnetic resonance imaging in subjects with prediabetes, diabetes, and normal glucose tolerance

Purpose

We hypothesize that MRI-based renal compartment volumes, particularly renal sinus fat as locally and potentially independently acting perivascular fat tissue, increase with glucose intolerance. We therefore analyze the distribution of renal volumes in individuals with normal glucose levels and prediabetic and diabetic individuals and investigate potential associations with other typical cardiometabolic biomarkers.

Material and methods

The sample comprised N = 366 participants who were either normoglycemic (N = 230), had prediabetes (N = 87) or diabetes (N = 49), as determined by Oral Glucose Tolerance Test. Other covariates were obtained by standardized measurements and interviews. Whole-body MR measurements were performed on a 3 Tesla scanner. For assessment of the kidneys, a coronal T1w dual-echo Dixon and a coronal T2w single shot fast spin echo sequence were employed. Stepwise semi-automated segmentation of the kidneys on the Dixon-sequences was based on thresholding and geometric assumptions generating volumes for the kidneys and sinus fat. Inter- and intra-reader variability were determined on a subset of 40 subjects. Associations between glycemic status and renal volumes were evaluated by linear regression models, adjusted for other potential confounding variables. Furthermore, the association of renal volumes with visceral adipose tissue was assessed by linear regression models and Pearson’s correlation coefficient.

Results

Renal volume, renal sinus volume and renal sinus fat increased gradually from normoglycemic controls to individuals with prediabetes to individuals with diabetes (renal volume: 280.3±64.7 ml vs 303.7±67.4 ml vs 320.6±77.7ml, respectively, p < 0.001). After adjustment for age and sex, prediabetes and diabetes were significantly associated to increased renal volume, sinus volume (e.g. β_Prediabetes = 10.1, 95% CI: [6.5, 13.7]; p<0.01, β_Diabetes = 11.86, 95% CI: [7.2, 16.5]; p<0.01) and sinus fat (e.g. β_Prediabetes = 7.13, 95% CI: [4.5, 9.8]; p<0.001, β_Diabetes = 7.34, 95% CI: [4.0, 10.7]; p<0.001). Associations attenuated after adjustment for additional confounders were only significant for prediabetes and sinus volume (ß = 4.0 95% CI [0.4, 7.6]; p<0.05). Hypertension was significantly associated with increased sinus volume (β = 3.7, 95% CI: [0.4, 7.0; p<0.05]) and absolute sinus fat volume (β = 3.0, 95% CI: [0.7, 5.3]; p<0.05). GFR and all renal volumes were significantly associated as well as urine creatinine levels and renal sinus volume (β = 1.6, 95% CI: [0.1, 2.9]; p<0.05).

Conclusion

Renal volume and particularly renal sinus fat volume already increases significantly in prediabetic subjects and is significantly associated with VAT. This shows, that renal sinus fat is a perivascular adipose tissue, which early undergoes changes in the development of metabolic disease. Our findings underpin that renal sinus fat is a link between metabolic disease and associated chronic kidney disease, making it a potential imaging biomarker when assessing perivascular adipose tissue.

Strategic research agenda for biomedical imaging

This Strategic Research Agenda identifies current challenges and needs in healthcare, illustrates how biomedical imaging and derived data can help to address these, and aims to stimulate dedicated research funding efforts.Medicine is currently moving towards a more tailored, patient-centric approach by providing personalised solutions for the individual patient. Innovation in biomedical imaging plays a key role in this process as it addresses the current needs for individualised prevention, treatment, therapy response monitoring, and image-guided surgery.The use of non-invasive biomarkers facilitates better therapy prediction and monitoring, leading to improved patient outcomes. Innovative diagnostic imaging technologies provide information about disease characteristics which, coupled with biological, genetic and -omics data, will contribute to an individualised diagnosis and therapy approach.In the emerging field of theranostics, imaging tools together with therapeutic agents enable the selection of best treatments and allow tailored therapeutic interventions.For prenatal monitoring, the use of innovative imaging technologies can ensure an early detection of malfunctions or disease.The application of biomedical imaging for diagnosis and management of lifestyle-induced diseases will help to avoid disease development through lifestyle changes.Artificial intelligence and machine learning in imaging will facilitate the improvement of image interpretation and lead to better disease prediction and therapy planning.As biomedical imaging technologies and analysis of existing imaging data provide solutions to current challenges and needs in healthcare, appropriate funding for dedicated research is needed to implement the innovative approaches for the wellbeing of citizens and patients.