Automated detection of enteric tubes misplaced in the respiratory tract on chest radiographs using deep learning with two centre validation.

AIM: To develop and test a model based on a convolutional neural network that can identify enteric tube position accurately on chest radiography. MATERIALS AND METHODS: The chest radiographs of adult patients were classified by radiologists based on enteric tube position as either critically misplaced (within the respiratory tract) or not critically misplaced (misplaced within the oesophagus or safely positioned below the diaphragm). A deep-learning model based on the 121-layer DenseNet architecture was developed using a training and validation set of 4,693 chest radiographs. The model was evaluated on an external test data set from a separate institution that consisted of 1,514 consecutive radiographs with a real-world incidence of critically misplaced enteric tubes. RESULTS: The receiver operator characteristic area under the curve was 0.90 and 0.92 for the internal validation and external test data sets, respectively. For the external data set with a prevalence of 4.4% of critically misplaced enteric tubes, the model achieved high accuracy (92%), sensitivity (80%), and specificity (92%) for identifying a critically misplaced enteric tube. The negative predictive value (99%) was higher than the positive predictive value (32%). CONCLUSION: The present study describes the development and external testing of a model that accurately identifies an enteric tube misplaced within the respiratory tract. This model could help reduce the risk of the catastrophic consequences of feeding through a misplaced enteric tube.

Putting machine learning into motion: applications in cardiovascular imaging.

Heart and circulatory diseases cause a quarter of all deaths in the UK and cardiac imaging offers an effective tool for early diagnosis and risk-stratification to improve premature death and disability. This domain of radiology is unique in that assessing flow and motion is essential for understanding and quantifying normal physiology and disease processes. Conventional image interpretation relies on manual analysis but this often fails to capture important prognostic features in the complex disturbances of cardiovascular physiology. Machine learning (ML) in cardiovascular imaging promises to be a transformative tool and addresses an unmet need for patient-specific management, accurate prediction of future events, and the discovery of tractable molecular mechanisms of disease. This review discusses the potential of ML across every aspect of image analysis including efficient acquisition, segmentation and motion tracking, disease classification, prediction tasks and modelling of genotype-phenotype interactions; however, significant challenges remain in access to high-quality data at scale, robust validation, and clinical interpretability.

Identifying the optimal regional predictor of right ventricular global function: a high-resolution three-dimensional cardiac magnetic resonance study.

Right ventricular (RV) function has prognostic value in acute, chronic and peri-operative disease, although the complex RV contractile pattern makes rapid assessment difficult. Several two-dimensional (2D) regional measures estimate RV function, however the optimal measure is not known. High-resolution three-dimensional (3D) cardiac magnetic resonance cine imaging was acquired in 300 healthy volunteers and a computational model of RV motion created. Points where regional function was significantly associated with global function were identified and a 2D, optimised single-point marker (SPM-O) of global function developed. This marker was prospectively compared with tricuspid annular plane systolic excursion (TAPSE), septum-freewall displacement (SFD) and their fractional change (TAPSE-F, SFD-F) in a test cohort of 300 patients in the prediction of RV ejection fraction. RV ejection fraction was significantly associated with systolic function in a contiguous 7.3 cm2 patch of the basal RV freewall combining transverse (38%), longitudinal (35%) and circumferential (27%) contraction and coinciding with the four-chamber view. In the test cohort, all single-point surrogates correlated with RV ejection fraction (p < 0.010), but correlation (R) was higher for SPM-O (R = 0.44, p < 0.001) than TAPSE (R = 0.24, p < 0.001) and SFD (R = 0.22, p < 0.001), and non-significantly higher than TAPSE-F (R = 0.40, p < 0.001) and SFD-F (R = 0.43, p < 0.001). SPM-O explained more of the observed variance in RV ejection fraction (19%) and predicted it more accurately than any other 2D marker (median error 2.8 ml vs 3.6 ml, p < 0.001). We conclude that systolic motion of the basal RV freewall predicts global function more accurately than other 2D estimators. However, no markers summarise 3D contractile patterns, limiting their predictive accuracy.

Computational study of aortic hemodynamics for patients with an abnormal aortic valve: The importance of secondary flow at the ascending aorta inlet.

Blood flow in the aorta is helical, but most computational studies ignore the presence of secondary flow components at the ascending aorta (AAo) inlet. The aim of this study is to ascertain the importance of inlet boundary conditions (BCs) in computational analysis of flow patterns in the thoracic aorta based on patient-specific images, with a particular focus on patients with an abnormal aortic valve. Two cases were studied: one presenting a severe aortic valve stenosis and the other with a mechanical valve. For both aorta models, three inlet BCs were compared; these included the flat profile and 1D through-plane velocity and 3D phase-contrast magnetic resonance imaging derived velocity profiles, with the latter being used for benchmarking. Our results showed that peak and mean velocities at the proximal end of the ascending aorta were underestimated by up to 41% when the secondary flow components were neglected. The results for helical flow descriptors highlighted the strong influence of secondary velocities on the helical flow structure in the AAo. Differences in all wall shear stress (WSS)-derived indices were much more pronounced in the AAo and aortic arch (AA) than in the descending aorta (DAo). Overall, this study demonstrates that using 3D velocity profiles as inlet BC is essential for patient-specific analysis of hemodynamics and WSS in the AAo and AA in the presence of an abnormal aortic valve. However, predicted flow in the DAo is less sensitive to the secondary velocities imposed at the inlet; hence, the 1D through-plane profile could be a sufficient inlet BC for studies focusing on distal regions of the thoracic aorta.

Diet-induced obesity causes visceral, but not subcutaneous, lymph node hyperplasia via increases in specific immune cell populations.

OBJECTIVES: The spatial proximity of adipose depots to secondary lymph nodes allows a unique relation between the two systems. Obesity, predominately visceral adiposity, links to numerous diseases; hence, we postulate that secondary lymphatics within this region contributes to disease risk. MATERIAL AND METHODS: Male C57BL/6 mice were fed standard CHOW (18% kcal fat) or Western diet (45% kcal fat) for 7 weeks. Visceral and subcutaneous lymph nodes and associated adipose depots they occupy were excised. Lymph node morphology and resident immune cell populations were characterized via histopathology, immunofluorescence and flow cytometry. Adipose tissue immune cell populations were also characterized. RESULTS: Obesity caused lymph node expansion, increased viable cell number and deviations in immune cell populations. These alterations were exclusive to visceral lymph nodes. Notably, pro-inflammatory antigen presenting cells and regulatory T cells increased in number in the visceral lymph node. Obesity, however, reduced T regulatory cells in visceral lymph nodes. The visceral adipose depot also had greater reactivity towards HFD than subcutaneous, with a greater percent of macrophages, dendritic and CD8+ T cells. Immune cell number, in both the visceral and subcutaneous, however decreased as adipose depots enlarged. CONCLUSION: Overall, HFD has a greater influence on visceral cavity than the subcutaneous. In the visceral lymph node, but not subcutaneous, HFD-induced obesity decreased cell populations that suppressed immune function while increasing those that regulate/activate immune response.

On the choice of outlet boundary conditions for patient-specific analysis of aortic flow using computational fluid dynamics.

Boundary conditions (BCs) are an essential part in computational fluid dynamics (CFD) simulations of blood flow in large arteries. Although several studies have investigated the influence of BCs on predicted flow patterns and hemodynamic wall parameters in various arterial models, there is a lack of comprehensive assessment of outlet BCs for patient-specific analysis of aortic flow. In this study, five different sets of outlet BCs were tested and compared using a subject-specific model of a normal aorta. Phase-contrast magnetic resonance imaging (PC-MRI) was performed on the same subject and velocity profiles extracted from the in vivo measurements were used as the inlet boundary condition. Computational results obtained with different outlet BCs were assessed in terms of their agreement with the PC-MRI velocity data and key hemodynamic parameters, such as pressure and flow waveforms and wall shear stress related indices. Our results showed that the best overall performance was achieved by using a well-tuned three-element Windkessel model at all model outlets, which not only gave a good agreement with in vivo flow data, but also produced physiological pressure waveforms and values. On the other hand, opening outlet BCs with zero pressure at multiple outlets failed to reproduce any physiologically relevant flow and pressure features.

Role of monocyte recruitment in hemangiosarcoma metastasis in dogs.

Canine hemangiosarcoma (HSA) is a highly malignant tumour associated with short survival times because of early and widespread metastasis. In humans and rodents, monocytes play key roles in promoting tumour metastasis through stimulating tumour cell extravasation, seeding, growth and angiogenesis. Therefore, we investigated the potential association between monocyte infiltration and tumour metastasis in HSA and other common canine tumours. Immunohistochemistry was used to quantify CD18+ monocytes within metastases. We found that HSA metastases had significantly greater numbers of CD18+ monocytes compared with metastases from other tumour types. HSA cells were the highest producers of the monocyte chemokine CCL2, and stimulated canine monocyte migration in a CCL2 dependent manner. These results are consistent with the hypothesis that overexpression of CCL2 and recruitment of large numbers of monocytes may explain in part the aggressive metastatic nature of canine HSA. Thus, therapies designed to block monocyte recruitment may be an effective adjuvant strategy for suppressing HSA metastasis in dogs.

Acute myocardial infarction: susceptibility-weighted cardiac MRI for the detection of reperfusion haemorrhage at 1.5 T.

AIM: To assess whether susceptibility-weighted imaging (SWI) provides better image contrast for the detection of haemorrhagic ischaemia-reperfusion injury in the heart. MATERIALS AND METHODS: Thirty patients (all men; mean age 53 years) underwent cardiac magnetic resonance imaging (MRI) within 7 days of primary percutaneous intervention for acute ST elevation myocardial infarction (STEMI). Multiple gradient-echo T2* sequences with magnitude and phase reconstructions were acquired. A high-pass filtered phase map was used to create a mask for the SWI reconstructions. The difference in image contrast was assessed in those patients with microvascular obstruction. A mixed effects regression model was used to test the effect of echo time and reconstruction method on phase and contrast-to-noise ratio (CNR). Medians and interquartile ranges (IQR) are reported. RESULTS: T2* in haemorrhagic infarcts was shorter than in non-haemorrhagic infarcts (33.5 ms [24.9-43] versus 49.9 ms [44.6-67.6]; p=0.0007). The effect of echo time on phase was significant (p<0.0001), as was the effect of haemorrhage on phase (p=0.0016). SWI reconstruction had a significant effect on the CNR at all echo times (echoes 1-5, p<0.0001; echo 6, p=0.01; echo 7, p=0.02). The median echo number at which haemorrhage was first visible was less for SWI compared to source images (echo 2 versus echo 5, p=0.0002). CONCLUSION: Cardiac SWI improves the contrast between myocardial haemorrhage and the surrounding tissue following STEMI and has potential as a new tool for identifying patients with ischaemia-reperfusion injury.

White matter brain lesions in midlife familial hypercholesterolemic patients at 3-Tesla magnetic resonance imaging.

BACKGROUND: Patients with hypercholesterolemia of 60 years and older have an increased risk for white matter brain lesions and dementia. PURPOSE: To investigate whether patients with familial hypercholesterolemia (FH) develop white matter lesions at 3-Tesla (T) MRI as early as in midlife. MATERIAL AND METHODS: Non-diabetic, nonsmoking, and non-hypertensive heterozygous FH patients on treatment with maximally tolerated dose of a statin for more than 5 years (n = 14) and matched controls (n = 22) aged 25 to 60 years of age were studied. Imaging was performed at 3T with a fluid-attenuated T2-weighted MR pulse sequence and a T1-weighted spin-echo pulse sequence following 10 ml of i.v. gadopentetate dimeglumine. Images were evaluated by two independent readers. Fasting blood samples were taken. Student's t test was employed at P<0.05. RESULTS: Three volunteers and one FH patient had white matter lesions (P<0.53). No other evidence of past ischemic stroke was observed. Mean total serum cholesterol and low-density lipoprotein (LDL) cholesterol were significantly higher in the FH group (6.0+/-1.1 vs. 5.1+/-0.9 mmol/l, P<0.02 and 4.1+/-0.9 vs. 3.1+/-0.8 mmol/l, P<0.004, respectively). CONCLUSION: Heterozygous FH patients on statin treatment in the age range of 25 to 60 years are not at increased risk of white matter lesions at 3T MRI.

Establishing a clinical cardiac MRI service.

After several years of research development cardiovascular MRI has evolved into a widely accepted clinical tool. It offers important diagnostic and prognostic information for a variety of clinical indications, which include ischaemic heart disease, cardiomyopathies, valvular dysfunction and congenital heart disorders. It is a safe non-invasive technique that employs a variety of imaging sequences optimized for temporal or spatial resolution, tissue-specific contrast, flow quantification or angiography. Cardiac MRI offers specific advantages over conventional imaging techniques for a significant number of patients. The demand for cardiac MRI studies from cardiothoracic surgeons, cardiologists and other referrers is likely to continue to rise with pressure for more widespread local service provision. Setting up a cardiac MRI service requires careful consideration regarding funding issues and how it will be integrated with existing service provision. The purchase of cardiac phased array coils, monitoring equipment and software upgrades must also be considered, as well as the training needs of those involved. The choice of appropriate imaging protocols will be guided by operator experience, clinical indication and equipment capability, and is likely to evolve as the service develops. Post-processing and offline analysis form a significant part of the time taken to report studies and an efficient method of providing quantitative reports is an important requirement. Collaboration between radiologists and cardiologists is needed to develop a successful service and multi-disciplinary meetings are key component of this. This review will explore these issues from our perspective of a new clinical cardiac MRI service operating over its first year in a teaching hospital imaging department.