Detection and quantification of breast arterial calcifications on mammograms: a deep learning approach.

OBJECTIVE: Breast arterial calcifications (BAC) are a sex-specific cardiovascular disease biomarker that might improve cardiovascular risk stratification in women. We implemented a deep convolutional neural network for automatic BAC detection and quantification. METHODS: In this retrospective study, four readers labelled four-view mammograms as BAC positive (BAC+) or BAC negative (BAC-) at image level. Starting from a pretrained VGG16 model, we trained a convolutional neural network to discriminate BAC+ and BAC- mammograms. Accuracy, F1 score, and area under the receiver operating characteristic curve (AUC-ROC) were used to assess the diagnostic performance. Predictions of calcified areas were generated using the generalized gradient-weighted class activation mapping (Grad-CAM++) method, and their correlation with manual measurement of BAC length in a subset of cases was assessed using Spearman ρ. RESULTS: A total 1493 women (198 BAC+) with a median age of 59 years (interquartile range 52-68) were included and partitioned in a training set of 410 cases (1640 views, 398 BAC+), validation set of 222 cases (888 views, 89 BAC+), and test set of 229 cases (916 views, 94 BAC+). The accuracy, F1 score, and AUC-ROC were 0.94, 0.86, and 0.98 in the training set; 0.96, 0.74, and 0.96 in the validation set; and 0.97, 0.80, and 0.95 in the test set, respectively. In 112 analyzed views, the Grad-CAM++ predictions displayed a strong correlation with BAC measured length (ρ = 0.88, p < 0.001). CONCLUSION: Our model showed promising performances in BAC detection and in quantification of BAC burden, showing a strong correlation with manual measurements. CLINICAL RELEVANCE STATEMENT: Integrating our model to clinical practice could improve BAC reporting without increasing clinical workload, facilitating large-scale studies on the impact of BAC as a biomarker of cardiovascular risk, raising awareness on women's cardiovascular health, and leveraging mammographic screening. KEY POINTS: • We implemented a deep convolutional neural network (CNN) for BAC detection and quantification. • Our CNN had an area under the receiving operator curve of 0.95 for BAC detection in the test set composed of 916 views, 94 of which were BAC+ . • Furthermore, our CNN showed a strong correlation with manual BAC measurements (ρ = 0.88) in a set of 112 views.

Mammography biomarkers of cardiovascular and musculoskeletal health: A review.

Breast density (BD) and breast arterial calcifications (BAC) can expand the role of mammography. In premenopause, BD is related to body fat composition: breast adipose tissue and total volume are potential indicators of fat storage in visceral depots, associated with higher risk of cardiovascular disease (CVD). Women with fatty breast have an increased likelihood of hypercholesterolemia. Women without cardiometabolic diseases with higher BD have a lower risk of diabetes mellitus, hypertension, chest pain, and peripheral vascular disease, while those with lower BD are at increased risk of cardiometabolic diseases. BAC, the expression of Monckeberg sclerosis, are associated with CVD risk. Their prevalence, 13 % overall, rises after menopause and is reduced in women aged over 65 receiving hormonal replacement therapy. Due to their distinct pathogenesis, BAC are associated with hypertension but not with other cardiovascular risk factors. Women with BAC have an increased risk of acute myocardial infarction, ischemic stroke, and CVD death; furthermore, moderate to severe BAC load is associated with coronary artery disease. The clinical use of BAC assessment is limited by their time-consuming manual/visual quantification, an issue possibly solved by artificial intelligence-based approaches addressing BAC complex topology as well as their large spectrum of extent and x-ray attenuations. A link between BD, BAC, and osteoporosis has been reported, but data are still inconclusive. Systematic, standardised reporting of BD and BAC should be encouraged.

Quantitative Assessment of Late Gadolinium Enhancement and Edema at Cardiac Magnetic Resonance in Low-Risk Myocarditis Patients.

In this study, we aimed to quantify LGE and edema at short-tau inversion recovery sequences on cardiac magnetic resonance (CMR) in patients with myocarditis. We retrospectively evaluated CMR examinations performed during the acute phase and at follow-up. Forty-seven patients were eligible for retrospective LGE assessment, and, among them, twenty-five patients were eligible for edema evaluation. Both groups were paired with age- and sex-matched controls. The median left ventricle LGE was 6.4% (interquartile range 5.0−9.2%) at the acute phase, 4.4% (3.3−7.2%) at follow-up, and 4.3% (3.0−5.3%) in controls, the acute phase being higher than both follow-up and controls (p < 0.001 for both), while follow-up and controls did not differ (p = 0.139). An optimal threshold of 5.0% was obtained for LGE with 87% sensitivity and 48% specificity; the positive likelihood ratio (LR) was 1.67, and the negative LR was 0.27. Edema was 12.8% (9.4−18.1%) at the acute phase, 7.3% (5.5−8.8%) at follow-up, and 6.7% (5.6−8.6%) in controls, the acute phase being higher than both follow-up and controls (both p < 0.001), while follow-up and controls did not differ (p = 0.900). An optimal threshold of 9.5% was obtained for edema with a sensitivity of 76% and a specificity of 88%; the positive LR was 6.33, and the negative LR was 0.27. LGE and edema thresholds are useful in cases of suspected mild myocarditis.

Safe Follow-Up after Endovascular Aortic Repair with Unenhanced MRI: The SAFEVAR Study.

We aimed to investigate whether unenhanced magnetic resonance imaging (MRI) could represent a safe and highly sensitive tool for endoleak screening in patients treated with endovascular aneurysm repair (EVAR) using computed tomography angiography (CTA) as a reference standard. Patients who underwent CTA for EVAR follow-up at our institution were prospectively enrolled. All MRI examinations were performed with a 1.5 T unit. The true-FISP and HASTE sequences of the MRI scans were assessed for the presence of hyperintensity within the aneurysm sac outside the graft, whereas phase-contrast through-plane sequences were used for blood flow quantification. We included 45 patients, 5 (11%) of whom were female. The median age was 73 years (IQR 68−78 years). Among our patients, 19 (42%) were positive for endoleaks at CTA, of whom 13 (68%) had type II endoleaks and 6 (32%) had type I endoleaks. There were no significant differences in age, sex, aneurysm type, prosthesis type, or contrast-to-noise ratio between hyperintensity and thrombus between patients with and without endoleaks (p > 0.300). The combined evaluation of true-FISP and HASTE yielded 100% sensitivity (95% CI: 79−100%) and 19% specificity (95% CI: 7−40%). Patients with a positive CTA had a median thrombus flow of 0.06 L/min (IQR 0.03−0.23 L/min), significantly greater than that of patients with a negative CTA (p = 0.007). Setting a threshold at 0.01 L/min, our MRI protocol yielded 100% sensitivity, 56% specificity, and an AUC of 0.76 (95% CI 0.60−0.91). In conclusion, unenhanced MRI has perfect sensitivity for endoleak detection, although with subpar specificity that could be improved with phase-contrast flow analysis.

Liquid Calibration Phantoms in Ultra-Low-Dose QCT for the Assessment of Bone Mineral Density.

INTRODUCTION: Cortical bone is affected by metabolic diseases. Some studies have shown that lower cortical bone mineral density (BMD) is related to increases in fracture risk which could be diagnosed by quantitative computed tomography (QCT). Nowadays, hybrid iterative reconstruction-based (HIR) computed tomography (CT) could be helpful to quantify the peripheral bone tissue. A key focus of this paper is to evaluate liquid calibration phantoms for BMD quantification in the tibia and under hybrid iterative reconstruction-based-CT with the different hydrogen dipotassium phosphate (K2HPO4) concentrations phantoms. METHODOLOGY: Four ranges of concentrations of K2HPO4 were made and tested with 2 exposure settings. Accuracy of the phantoms with ash gravimetry and intermediate K2HPO4 concentration as hypothetical patients were evaluated. The correlations and mean differences between measured equivalent QCT BMD and ash density as a gold standard were calculated. Relative percentage error (RPE) in CT numbers of each concentration over a 6-mo period was reported. RESULTS: The correlation values (R2 was close to 1.0), suggested that the precision of QCT-BMD measurements using standard and ultra-low dose settings were similar for all phantoms. The mean differences between QCT-BMD and the ash density for low concentrations (about 93 mg/cm3) were lower than high concentration phantoms with 135 and 234 mg/cm3 biases. In regard to accuracy test for hypothetical patient, RPE was up to 16.1% for the low concentration (LC) phantom for the case of high mineral content. However, the lowest RPE (0.4 to 1.8%) was obtained for the high concentration (HC) phantom, particularly for the high mineral content case. In addition, over 6 months, the K2HPO4 concentrations increased 25% for 50 mg/cm3 solution and 0.7 % for 1300 mg/cm3 solution in phantoms. CONCLUSION: The excellent linear correlations between the QCT equivalent density and the ash density gold standard indicate that QCT can be used with submilisivert radiation dose. We conclude that using liquid calibration phantoms with a range of mineral content similar to that being measured will minimize bias. Finally, we suggest performing BMD measurements with ultra-low dose scan concurrent with iterative-based reconstruction to reduce radiation exposure.

A hybrid (iron-fat-water) phantom for liver iron overload quantification in the presence of contaminating fat using magnetic resonance imaging.

OBJECTIVE: Assessment of iron content in the liver is crucial for diagnosis/treatment of iron-overload diseases. Nonetheless, T2*-based methods become challenging when fat and iron are simultaneously present. This study proposes a phantom design concomitantly containing various concentrations of iron and fat suitable for devising accurate simultaneous T2* and fat quantification technique. MATERIALS AND METHODS: A 46-vial iron-fat-water phantom with various iron concentrations covering clinically relevant T2* relaxation time values, from healthy to severely overloaded liver and wide fat percentages ranges from 0 to 100% was prepared. The phantom was constructed using insoluble iron (II, III) oxide powder containing microscale particles. T2*-weighted imaging using multi-gradient-echo (mGRE) sequence, and chemical shift imaging spin-echo (CSI-SE) Magnetic Resonance Spectroscopy (MRS) data were considered for the analysis. T2* relaxation times and fat fractions were extracted from the MR signals to explore the effects of fat and iron overload. RESULTS: Size distribution of iron oxide particles for Magnetite fits with a lognormal function with a mean size of about 1.17 µm. Comparison of FF color maps, estimated from bi- and mono-exponential model indicated that single-T2* fitting model resulted in lower NRMSD. Therefore, T2* values from the mono-exponential signal equation were used and expressed the relationship between relaxation time value across all iron (Fe) and fat concentration as [Formula: see text], with R-squared = 0.89. DISCUSSION: The proposed phantom design with microsphere iron particles closely simulated the single-T2* behavior of fatty iron-overloaded liver in vivo.