Deep learning performance compared to healthcare experts in detecting wrist fractures from radiographs: A systematic review and meta-analysis.

OBJECTIVE: To perform a systematic review and meta-analysis of the diagnostic accuracy of deep learning (DL) algorithms in the diagnosis of wrist fractures (WF) on plain wrist radiographs, taking healthcare experts consensus as reference standard. METHODS: Embase, Medline, PubMed, Scopus and Web of Science were searched in the period from 1 Jan 2012 to 9 March 2023. Eligible studies were patients with wrist radiographs for radial and ulnar fractures as the target condition, studies using DL algorithms based on convolutional neural networks (CNN), and healthcare experts consensus as the minimum reference standard. Studies were assessed with a modified QUADAS-2 tool, and we applied a bivariate random-effects model for meta-analysis of diagnostic test accuracy data. RESULTS: Our study was registered at PROSPERO with ID: CRD42023431398. We included 6 unique studies for meta-analysis, with a total of 33,026 radiographs. CNN performance compared to reference standards for the included articles found a summary sensitivity of 92% (95% CI: 80%-97%) and a summary specificity of 93% (95% CI: 76%-98%). The generalized bivariate I-squared statistic indicated considerable heterogeneity between the studies (81.90%). Four studies had one or more domains at high risk of bias and two studies had concerns regarding applicability. CONCLUSION: The diagnostic accuracy of CNNs was comparable to that of healthcare experts in wrist radiographs for investigation of WF. There is a need for studies with a robust reference standard, external data-set validation and investigation of diagnostic performance of healthcare experts aided with CNNs. CLINICAL RELEVANCE STATEMENT: DL matches healthcare experts in diagnosing WFs, which potentially benefits patient diagnosis.

Low-cost 3D-printed anthropomorphic cardiac phantom, for computed tomography automatic left ventricle segmentation and volumetry - A pilot study.

INTRODUCTION: Accurate cardiac left ventricle (LV) delineation is essential to CT-derived left ventricular ejection fraction (LVEF). To evaluate dose-reduction potential, an anatomically accurate heart phantom, with realistic X-ray attenuation is required. We demonstrated and tested a custom-made phantom using 3D-printing, and examined the influence of image noise on automatically measured LV volumes METHODS: A single coronary CT angiography (CCTA) dataset was segmented and converted to Standard Tessellation Language (STL) mesh, using open-source software. A 3D-printed model, with hollow left heart chambers, was printed and cavities filled with gelatinized contrast media. This was CT-scanned in an anthropomorphic chest phantom, at different exposure conditions. LV and "myocardium" noise and attenuation was measured. LV volume was automatically measured using two different methods. We calculated Spearmans' correlation of LV volume with noise and contrast-noise ratio respectively om 486 scans of the phantom. Source images were compared to one phantom series with similar parameters. This was done using Dice coefficient on LV short-axis segmentations. RESULTS: Phantom "Myocardium" and LV attenuation was comparable to measurements on source images. Automatic volume measurement succeeded, with mean volume deviation to patient images less than 2 ml. There was a moderate correlation of volume with CNR, and strong correlation of volume with image noise. With papillary muscles included in LV volume, the correlation was positive, but negative when excluded. Variation of volumes was lowest at 90-100 kVp for both methods in the 486 repeat scans. The Dice coefficient was 0.87, indicating high overlap between the single phantom series and source scan. Cost of 3D-printer and materials was 400 and 30 Euro respectively. CONCLUSION: Both anatomically and radiologically the phantom mimicked the source scans closely. LV volumetry was reliably performed with automatic algorithms. IMPLICATIONS FOR PRACTICE: Patient-specific cardiac phantoms may be produced at minimal cost and can potentially be used for other anatomies and pathologies. This enables radiographic phantom studies without need for dedicated 3D-labs or expensive commercial phantoms.

Amide proton transfer-weighted MRI in the clinical setting - correlation with dynamic susceptibility contrast perfusion in the post-treatment imaging of adult glioma patients at 3T.

INTRODUCTION: We investigated the correlation between amide proton transfer-weighted magnetic resonance imaging (APTw MRI) and dynamic susceptibility contrast (DSC) perfusion in order to assess the potential of APTw MRI as an alternative to DSC in adult brain tumor (glioma) imaging. METHODS: After Ethical Committee approval, forty adult patients, treated for histopathologically confirmed glioma (World Health Organization (WHO) grade II-IV), were prospectively imaged at 3 Tesla (3 T) with DSC perfusion and a commercially available three-dimensional (3D) APTw sequence. Two consultant neuroradiologists independently performed region of interest (ROI) measurements on relative cerebral blood volume (rCBV) and APTw maps, co-registered with anatomical images. The correlation APTw MRI-DSC perfusion was assessed using Spearman's rank-order test. Inter-observer agreement was evaluated by the intraclass correlation coefficient (ICC) and Bland-Altman (BA) plots. RESULTS: A statistically significant moderately strong positive correlation was observed between maximum rCBV (rCBVmax) and maximum APTw (APTwmax) values (observer 1: r = 0.73; p < 0.01; observer 2: r = 0.62; p < 0.01). We found good inter-observer agreement for APTwmax (ICC = 0.82; 95% confidence interval (CI) 0.66-0.90), with somewhat broad outer 95% CI for the BA Limits of Agreement (LoA) (-1.6 to 1.9). ICC for APTwmax was higher than ICC for rCBVmax (ICC = 0.74; 95%; CI 0.50-0.86), but the difference was not statistically significant. CONCLUSION: APTwmax values correlate positively with rCBVmax in patients treated for brain glioma. APTw imaging is a reproducible technique, with some observer dependence. Results need to be confirmed by a larger population analysis. IMPLICATIONS FOR PRACTICE: APTw MRI can be a useful addition to glioma follow-up imaging and a potential alternative to DSC perfusion, especially in patients where contrast agent is contraindicated.

Pulmonary arteriovenous malformations: a radiological and clinical investigation of 136 patients with long-term follow-up.

AIM: To assess the clinical outcome of patients with and without hereditary haemorrhagic telangiectasia (HHT) after embolisation of pulmonary arteriovenous malformations (PAVM) from a single national centre. MATERIALS AND METHODS: The present register-based observational study including all patients with PAVM treated with embolisation at a reference centre for HHT and PAVM was undertaken over a 20-year period. Demographic data, HHT genotyping, clinical presentation, and outcome were registered. Patients with HHT were compared to the patients without HHT. Clinical examination, contrast-enhanced echocardiography, and computed tomography (CT) were used to assess the clinical outcome at follow-up. RESULTS: One hundred and thirty-six patients with 339 PAVM underwent embolisation during the study period: 22 did not have HHT; 62% had HHT1, 10% had HHT2, 4% had JP-HHT, 8% had clinical HHT without identified genetic mutations. Solitary PAVM were more common among patients without HHT than with HHT. Mean follow-up after the first embolisation was 58 months. Mean age at first embolisation was 46.5 years, and at last follow-up 51.8 years. The clinical success without shunt at follow-up was 87%. The 30-day mortality related to the embolisation was 0%. Twenty patients died during follow-up (mean age 69 years). Most patients could be treated during one session, but many will need a long follow-up with repeated clinical examinations and embolisation. CONCLUSION: The majority of patients referred for embolisation of PAVM had HHT. Multiple PAVM is associated with HHT. Patients with PAVM should be screened for HHT and patients with HHT for PAVM. Embolisation is a safe procedure with high clinical success.

First experiences with model based iterative reconstructions influence on quantitative plaque volume and intensity measurements in coronary computed tomography angiography.

PURPOSE: Investigate the influence of adaptive statistical iterative reconstruction (ASIR) and the model-based IR (Veo) reconstruction algorithm in coronary computed tomography angiography (CCTA) images on quantitative measurements in coronary arteries for plaque volumes and intensities. METHODS: Three patients had three independent dose reduced CCTA performed and reconstructed with 30% ASIR (CTDIvol at 6.7 mGy), 60% ASIR (CTDIvol 4.3 mGy) and Veo (CTDIvol at 1.9 mGy). Coronary plaque analysis was performed for each measured CCTA volumes, plaque burden and intensities. RESULTS: Plaque volume and plaque burden show a decreasing tendency from ASIR to Veo as median volume for ASIR is 314 mm3 and 337 mm3-252 mm3 for Veo and plaque burden is 42% and 44% for ASIR to 39% for Veo. The lumen and vessel volume decrease slightly from 30% ASIR to 60% ASIR with 498 mm3-391 mm3 for lumen volume and vessel volume from 939 mm3 to 830 mm3. The intensities did not change overall between the different reconstructions for either lumen or plaque. CONCLUSION: We found a tendency of decreasing plaque volumes and plaque burden but no change in intensities with the use of low dose Veo CCTA (1.9 mGy) compared to dose reduced ASIR CCTA (6.7 mGy & 4.3 mGy), although more studies are warranted.

Contrast Enhanced Ultrasound in CT-undetermined Focal Liver Lesions.

Purpose: The purpose of this study was to examine the diagnostic accuracy of CEUS in the diagnosis of focal liver lesions, which were undetermined at CT scan. Materials and Methods: From January 2010 to December 2010, patients with CT-undetermined focal liver lesions were included in this study. A total of 78 patients were evaluated: 41 men and 37 women, mean age 61.8 years; age range 30-91 years. All patients were examined with GE LOGIQ 9E ultrasound scanners with contrast-specific software, and SonoVue intravenous bolus. The standard of reference was composite consisting of: percutaneous biopsy, surgical resection, PET/CT and clinical follow-up. Results: The 78 included patients had 163 undetermined focal liver lesions, mean size 1.1 cm, range 0.1-5.3 cm. There were 18 malignant and 145 benign liver lesions, as defined by the standard of reference. In differentiating between benign vs. malignant CEUS demonstrated sensitivity, specificity, PPV, NPV and accuracy of 94.4% (95% CI: 56.3-99.5%), 99.3% (95% CI: 94.9 -99.9%), 94.4% (95% CI: 56.3-99.5%), 99.3% (95% CI: 94.9-99.9%) and 98.7% (95% CI: 94.9-99.7%), respectively. If the CEUS-inconclusive results were assumed to indicate malignancy, then sensitivity, specificity, PPV, NPV and accuracy would be 95.8% (95% CI: 66.4-99.6%), 98.6% (95% CI: 94.4-99.7%), 92.0% (95CI: 65.1-98.6%), 99.3% (95% CI: 95.0-99.9%), 98.2% (95% CI: 94.4-99.5%). Conclusion: The results of this study showed a high diagnostic accuracy for CEUS in undetermined focal liver lesions found by CT. Due to a high diagnostic accuracy of CEUS in this study, it was cautiously concluded that CEUS is useful in differentiating between malignant and benign focal liver lesions in clinical practice.

Uncontrolled hypertension is associated with coronary artery calcification and electrocardiographic left ventricular hypertrophy: a case-control study.

We conducted a 1:2 matched case-control study in order to evaluate whether the prevalence of coronary artery calcium (CAC) and electrocardiographic left ventricular hypertrophy (LVH) or strain was higher in patients with uncontrolled hypertension than in subjects from the general population, and evaluate the association between CAC and LVH in patients with uncontrolled hypertension. Cases were patients with uncontrolled hypertension, whereas the controls were random individuals from the general population without cardiovascular disease. CAC score was assessed using a non-contrast computed tomographic scan. LVH was evaluated using the Sokolow-Lyon voltage combination and Cornell voltage-duration product, respectively. Associations between CAC, LVH and traditional cardiovascular risk factors were tested by means of ordinal, conditional and classic binary logistic regression models. We found that uncontrolled hypertension was independently associated with both an ordinal CAC score category (odds ratio (OR) 3.9 (95% CI, 1.6-9.1), P = 0.002), the presence of CAC score>99 (OR 4.5 (95% CI, 1.4-14.7), P = 0.01) and electrocardiographic LVH (OR 10.1 (95% CI, 3.4-30.2), P < 0.001) on both univariate and multivariable analyses. There was, however, no correlation between CAC and LVH. The lack of an association between CAC and LVH suggests that they are markers of different complications of hypertension and may have independent predictive values. Patients with both CAC and LVH may be at higher risk than those in whom only one of these markers is present.

Digital radiography: optimization of image quality and dose using multi-frequency software.

BACKGROUND: New developments in processing of digital radiographs (DR), including multi-frequency processing (MFP), allow optimization of image quality and radiation dose. This is particularly promising in children as they are believed to be more sensitive to ionizing radiation than adults. OBJECTIVE: To examine whether the use of MFP software reduces the radiation dose without compromising quality at DR of the femur in 5-year-old-equivalent anthropomorphic and technical phantoms. MATERIALS AND METHODS: A total of 110 images of an anthropomorphic phantom were imaged on a DR system (Canon DR with CXDI-50 C detector and MLT[S] software) and analyzed by three pediatric radiologists using Visual Grading Analysis. In addition, 3,500 images taken of a technical contrast-detail phantom (CDRAD 2.0) provide an objective image-quality assessment. RESULTS: Optimal image-quality was maintained at a dose reduction of 61% with MLT(S) optimized images. Even for images of diagnostic quality, MLT(S) provided a dose reduction of 88% as compared to the reference image. Software impact on image quality was found significant for dose (mAs), dynamic range dark region and frequency band. CONCLUSION: By optimizing image processing parameters, a significant dose reduction is possible without significant loss of image quality.

The relation between coronary artery calcification in asymptomatic subjects and both traditional risk factors and living in the city centre: a DanRisk substudy.

OBJECTIVE: To evaluate the association between the risk factor for living in the city centre as a surrogate for air pollution and the presence of coronary artery calcification (CAC) in a population of asymptomatic Danish subjects. DESIGN AND SUBJECTS: A random sample of 1825 men and women of either 50 or 60 years of age were invited to take part in a screening project designed to assess risk factors for cardiovascular disease (CVD). Noncontrast cardiac computed tomography was performed on all subjects, and their Agatston scores were calculated to evaluate the presence of subclinical coronary atherosclerosis. The relationship between CAC and several demographic and clinical parameters was evaluated using multivariate logistic regression. RESULTS: A total of 1225 individuals participated in the study, of whom 250 (20%) were living in the centres of major Danish cities. Gender and age showed the greatest association with the presence of CAC: the odds ratio (OR) for men compared with women was 3.2 [95% confidence interval (CI) 2.5-4.2; P < 0.0001], and the OR for subjects aged 60 versus those aged 50 years was 2.2 (95% CI 1.7-2.8; P < 0.0001). Other variables independently associated with the presence of CAC were diabetes and smoking with ORs of 2.0 (95% CI 1.1-3.5; P = 0.03) and 1.9 (95% CI 1.4-2.5, P < 0.0001), respectively. The adjusted OR for subjects living in city centres compared to those living outside was 1.8 (95% CI 1.3-2.4; P = 0.0003). CONCLUSION: Both conventional risk factors for CVD and living in a city centre are independently associated with the presence of CAC in asymptomatic middle-aged subjects.

Identification of asymptomatic type 2 diabetes mellitus patients with a low, intermediate and high risk of ischaemic heart disease: is there an algorithm?

AIMS/HYPOTHESIS: The leading cause of death in type 2 diabetes is cardiovascular disease (CVD). We examined the prevalence of myocardial ischaemia in type 2 diabetes patients and tried to establish an algorithm to identify patients with a high risk of ischaemic heart disease. METHODS: Type 2 diabetes patients who had no known or suspected CVD, and had been referred consecutively to a diabetes clinic for the first time (n=305; age 58.6+/-11.3 years; diabetes duration 4.5+/-5.3 years) were screened for myocardial ischaemia using myocardial perfusion scintigraphy (MPS). RESULTS: The univariate predictors of myocardial ischaemia were: atypical or typical angina pectoris, two or more traditional risk factors for CVD, BMI >32 kg/m2, systolic blood pressure >140 mmHg, HbA1c >8.5%, high-sensitivity C-reactive protein >4.0 mg/l, N-terminal pro-brain natriuretic peptide >300 pg/ml, left atrial volume index >32 ml/m2, left ventricular ejection fraction <50%, and carotid and peripheral arterial disease. The algorithm identified low (n=96), intermediate (n=65) and high risk groups (n=115), in which the prevalence of myocardial ischaemia was 15%,23% and 43%, respectively. Overall the algorithm reduced the number of patients referred to MPS from 305 to 144.However, the sensitivity and specificity of the algorithm was just 68% and 62%, respectively. CONCLUSIONS/INTERPRETATION: Our algorithm was able to stratify which patients had a low, intermediate or high risk of myocardial ischaemia based on MPS. However, the algorithm had low sensitivity and specificity, combined with high cost and time requirements. TRIAL REGISTRATION: clinicaltrials.gov NCT00298844 FUNDING: The study was funded by the Danish Cardio vascular Research Academy (DaCRA), The Danish Diabetes Association and The Danish Heart Foundation.

PET/CT in cancer: Methodological considerations for comparative diagnostic phase II studies with paired binary data.

OBJECTIVES: When the combined diagnostic imaging technique PET/CT is considered promising with respect to diagnosis/staging of a certain cancer type, a systematic investigation by means of clinical diagnostic studies in the target population is necessary to evaluate the usefulness of PET/CT compared to the current standard. It is often difficult to decide in advance whether it is appropriate to plan a superiority or non-inferiority study. We propose a statistical analysis strategy which is flexible enough to cope with both aims alike. METHODS: In opposition to clinical studies on drugs, each patient can be subjected to both PET/CT and the current standard, leading to paired observations of binary data (e.g., cancer = yes/no, stage = 0/1+). The analysis strategy focuses on point estimates and confidence intervals for the difference (or relative increase) in accuracy measures. RESULTS: Formulas for approximate 95% confidence intervals for the differences in sensitivity, specificity, positive and negative predictive values between PET/CT and the standard procedures are given, respectively. The strategy can also be applied if results obtained with a golden standard are not available in patients in whom both PET/CT and the standard procedure gave negative results. Sample sizes can and should be determined in an adaptive manner. CONCLUSIONS: Diagnostic studies to assess the merit of PET/CT in the diagnostic work-up of cancer patients can and should start with phase II studies focusing on 95% confidence intervals for differences in diagnostic measures. Even if the gold standard procedure is incomplete, the statistical analysis strategy given here may still be applicable.