Intraobserver Variability Research Articles

Echocardiographic indicators of fluid responsiveness in hospitalized dogs with compromised hemodynamics and tissue hypoperfusion.

To evaluate the accuracy of selected echocardiographic variables used to predict fluid responsiveness in hospitalized dogs with compromised hemodynamics and tissue hypoperfusion. Diagnostic test study in a prospective cohort of hospitalized dogs. Veterinary referral clinics. Forty-four hospitalized dogs with compromised hemodynamics and tissue hypoperfusion were utilized in this study. Echocardiographic examination before and after fluid replacement with 30ml/kg of lactated Ringer's solution. Pre-fluid replacement measurements of velocity of transmitral E wave (E-peak), the left ventricular end-diastolic internal diameter normalized to body weight (LVIDdN), and the left ventricular end-systolic internal diameter normalized to body weight (LVIDsN) were significantly lower in fluid-responsive patients compared with nonresponders (P<0.001). The area under the receiver operating characteristic curve (AUROC) with its 95% confidence interval (CI) for each significant predictor was as follows: E-peak 0.907 (0.776-1.000, P<0.001) and LVIDdN 0.919 (0.801-1.000, P<0.001). The predictive capacity of LVIDsN was not significantly better than chance (AUROC, 0.753; 95% CI, 0.472-1.000, P=0.078). A significant negative linear correlation was observed between the percentage of increase in velocity-time integral after expansion and the echocardiographic variables LVIDdN (rs =-0.452, P =0.023) and E-peak (rs =-0.396, P=0.008) pre-fluid replacement. The intraobserver and interobserver variability was very low(<5 %) for all measurements. In this study using critically ill dogs with compromised hemodynamics and tissue hypoperfusion, pre-fluid replacement measurements of LVIDdN and E-peak adequately predict fluid responsiveness. Because a small number of fluid nonresponders were involved in the present study (11.4%), further studies that include larger numbers of fluid-nonresponsive animals are required.

Clinical, Experimental, and Computational Validation of a New Doppler-Based Index for Coarctation Severity Assessment

Long-term morbidity including hypertension often persists in coarctation patients despite current guidelines. Coarctation severity can be invasively assessed via peak-to-peak catheter pressure gradient (PPCG), which is estimated noninvasively via simplified Bernoulli equation and conventionally reported as peak instantaneous Doppler gradient (PIDG). However, underlying simplifications of the equation limit diagnostic accuracy. We studied the diagnostic performance of a new Doppler-based diastolic index called the continuous flow pressure gradient (CFPG) versus conventional indices in assessing coarctation severity. In a rabbit model mimicking human aortic coarctation, temporal blood pressure waveforms revealed the diastolic instantaneous pressure gradients and spectral Doppler features impacted by coarctation severity. We therefore hypothesized that CFPG provides superior correlation with coarctation gradients measured invasively. PIDG and CFPG were quantified using color flow echocardiography in humans and rabbits with discrete coarctations. Results were compared with PPCG in rabbits (n=34) and arm-leg systolic gradients (n=25) in humans via 1-way analysis of variance, Pearson's correlation, linear regression, and Bland-Altman analysis. A threshold of CFPG ≥ 4.6mm Hg was identified via the Youden index as representative of PPCG ≥ 20mm Hg (the current guideline value for coarctation intervention) in rabbits, while a CFPG ≥1.0mm Hg represented an arm-leg systolic gradient ≥20mm Hg in humans. Accuracy measures revealed superior correlation of CFPG (R2>0.80) and mild receiver operating characteristic improvement (area under the receiver operating characteristic curve, 0.94-0.95) compared with PIDG (R2<0.63; area under the receiver operating characteristic curve, 0.89-0.95). Inter-/intraobserver variability tested by intraclass correlation coefficient revealed measurement reliability with differences ≤8.2% and 10.7%, respectively. Computational simulations of anesthetized versus conscious hemodynamics showed parameters were minimally impacted by isoflurane inherent in the dataused to derive CFPG. These results confirm the potential diagnostic accuracy of CFPG in echocardiography-based coarctation severity assessment. We are optimistic that CFPG will be useful for translation of results from preclinical studies that revisit current guidelines to limit morbidity in humans with aortic coarctation.

Association of CTG Diagnosis of Intrapartum Fetal Distress and Immediate Postpartum Acidemia in Foetal Umbilical Artery.

Intrapartum fetal hypoxia which is one of the leading causes of neonatal morbidity and mortality is a preventable cause. Over the past years, many methods have been employed to diagnose fetal distress, a sign of fetal hypoxia, among these, cardiotocography (CTG) is the most widely used method. Diagnosis of fetal distress based on CTG can have high inter and intraobserver variation leading to either delayed or inessential intervention henceforth increasing maternal morbidity and mortality. Fetal cord arterial blood pH is an objective method to diagnose intrapartum fetal hypoxia, hence by observing the incidence of acidemia in cord blood pH among those newborns born through cesarean section (CS) in view of non-reassuring CTG can help make a judicious decision. In this single institutional observational study, patients admitted for safe confinement were subjected to CTG in the latent and active phases of labor. Non-Reassuring traces were further classified based on NICE guideline CG190. The cord blood of these neonates born through CS in view of non-reassuring CTG was drawn and sent for ABG analysis. Among the 87 neonates delivered through CS in view of fetal distress, 19.5% had acidosis. Among those with pathological traces, 16(28.6%) had acidosis and one (100%) which was categorized as the need for urgent intervention showed acidosis. This result showed a statistically significant association (p value-0.003). No statistically significant association was obtained when variation in baseline characters of CTG when taken independently. In our study, neonatal acidemia which is the objective evidence of fetal distress was seen in 19.5% of our study population who underwent cesarean in view of non-reassuring CTG. Among these, acidemia was significantly associated with pathological CTG trace compared to suspicious trace. We also observed that abnormal FHR features when taken independently showed no significant association with acidosis. Acidosis among newborns certainly increased the requirement for active resuscitation and additional hospital stay. Hence, we conclude that by recognizing specific fetal heart rate patterns associated with acidosis in a fetus, a more judicious decision can be made, thereby preventing both delayed as well as inessential intervention.

Collaboration of features optimization techniques for the effective diagnosis of glaucoma in retinal fundus images

Glaucoma is the second most common cause of vision loss. Manual screening of a patient's eye or screening through a fundus image of the patient's eye requires expert ophthalmologists. This screening analysis is time-consuming, requires expert human involvement, and is subject to human intra-observer variability. Thus, medical imaging professionals are working to solve these issues by investigating retinal images for glaucoma detection using artificial intelligence-based computer-aided diagnosis systems (CAD). Machine learning algorithms (for classification) and nature-inspired computing (for feature selection/reduction) embedded CAD systems can successfully identify retinal pictures and can be employed to overcome these challenges. This proposed work is a productive attempt in which we have proposed two novel two-layered approaches (BA-BCS, BCS-PSO) which are based on Particle Swarm Optimization (PSO), Binary Cuckoo Search (BCS), and Bat Algorithm (BA). We have also analyzed the performances of BA, BCS, and PSO separately. These five (single and two-layered) approaches are used to compose subsets of reduced features that can generate the maximum accuracy when forwarded to three machine learning classifiers. Benchmark publicly available datasets, ORIGA and REFUGE, and their combinations are used to validate the proposed methodology. A maximum accuracy of up to 98.95% is achieved using these approaches. Apart from this, many other trade-off solutions are also suggested for the researcher's community. This study therefore presents novel efforts with new and efficient results that are beneficial to ophthalmologists, researchers, and humanity.

Electrocardiography and heart rate variability in Göttingen Minipigs: Impact of diurnal variation, lead placement, repeatability and streptozotocin-induced diabetes

BackgroundThe Göttingen Minipig is widely used in preclinical research and safety pharmacology, but standardisation of porcine electrocardiography (ECG) is lacking. The aim of this study was to investigate diurnal effects, change over time and choice of lead on ECG morphology and heart rate variability (HRV) in healthy and streptozotocin (STZ) induced diabetic Göttingen Minipigs. MethodsDiabetes was experimentally induced using STZ in 11 Göttingen Minipigs (DIA). Seven controls (CON) were included. 24-h ECG was recorded at baseline and four months. Morphological parameters (QRS and T wave duration, P- and T-wave amplitude, PR and QT (Bazett's (QTcb) or Fridericia (QTcf) correction) intervals and ST segment), presence of cardiac arrhythmias, heart rate (HR) and HRV (time and frequency domain) were analysed. ResultsFour months after induction, DIA had decreased P-wave amplitude (P < 0.0001) and T-wave duration (P = 0.017), compared to CON. QTcb was lower in DIA, but not in CON. Both groups had decreased HR (P < 0.0001) and QRS duration (lead II, P = 0.04) and length of PR-segment increased (lead I and II, P < 0.01) while selected HRV parameters also increased (all P < 0.01). Time of day influenced HR, QRS duration, PR segment, ST segment, T- and P-wave amplitude and some parameters of HRV. Inter- and intra-observer variability of morphological measurements was low (<6%). ConclusionECG parameters were influenced by time setting, diurnal variation and lead. Some ECG and HRV changes were found in diabetic minipigs four months after STZ induction. The findings underline the need for standardisation of ECG and HRV in Göttingen Minipigs.

Quantitative Three-Dimensional Measurements of Acetabular Fracture Displacement Could Be Predictive for Native Hip Survivorship.

This study aims to develop a three-dimensional (3D) measurement for acetabular fracture displacement, determine the inter- and intra-observer variability, and correlate the measurement with clinical outcome. Three-dimensional models were created for 100 patients surgically treated for acetabular fractures. The ‘3D gap area’, the 3D surface between all the fracture fragments, was developed. The association between the 3D gap area and the risk of conversion to a total hip arthroplasty (THA) was determined by an ROC curve and a Cox regression analysis. The 3D gap area had an excellent inter-observer and intra-observer reliability. The preoperative median 3D gap area for patients without and with a THA was 1731 mm2 versus 2237 mm2. The median postoperative 3D gap area was 640 mm2 versus 845 mm2. The area under the curve was 0.63. The Cox regression analysis showed that a preoperative 3D gap area > 2103 mm2 and a postoperative 3D gap area > 1058 mm2 were independently associated with a 3.0 versus 2.4 times higher risk of conversion to a THA. A 3D assessment of acetabular fractures is feasible, reproducible, and correlates with clinical outcome. Three-dimensional measurements could be added to the current classification systems to quantify the level of fracture displacement and to assess operative results.

Artificial intelligence based liver portal tract region identification and quantification with transplant biopsy whole-slide images

Liver fibrosis staging is clinically important for liver disease progression prediction. As the portal tract fibrotic quantity and size in a liver biopsy correlate with the fibrosis stage, an accurate analysis of portal tract regions is clinically critical. Manual annotations of portal tract regions, however, are time-consuming and subject to large inter- and intra-observer variability. To address such a challenge, we develop a Multiple Up-sampling and Spatial Attention guided UNet model (MUSA-UNet) to segment liver portal tract regions in whole-slide images of liver tissue slides. To enhance the segmentation performance, we propose to use depth-wise separable convolution, the spatial attention mechanism, the residual connection, and multiple up-sampling paths in the developed model. This study includes 53 histopathology whole slide images from patients who received liver transplantation. In total, 6,012 patches derived from 30 images are used for our deep learning model training and validation. The remaining 23 whole slide images are utilized for the model testing. The average liver portal tract segmentation performance of the developed MUSA-UNet is 0.94 (Precision), 0.85 (Recall), 0.89 (F1 Score), 0.89 (Accuracy), 0.80 (Jaccard Index), and 0.91 (Fowlkes–Mallows Index), respectively. The clinical Scheuer fibrosis stage presents a strong correlation with the resulting average portal tract fibrotic area (R = 0.681, p<0.001) and portal tract percentage (R = 0.335, p = 0.02) computed from the MUSA-UNet segmentation results. In conclusion, our developed deep learning model MUSA-UNet can accurately segment portal tract regions from whole-slide images of liver tissue biopsies, presenting its promising potential to assist liver disease diagnosis in a computational manner.

Label noise and self-learning label correction in cardiac abnormalities classification

Objective. Learning to classify cardiac abnormalities requires large and high-quality labeled datasets, which is a challenge in medical applications. Small datasets from various sources are often aggregated to meet this requirement, resulting in a final dataset prone to label noise due to inter- and intra-observer variability and different expertise. It is well known that label noise can affect the performance and generalizability of the trained models. In this work, we explore the impact of label noise and self-learning label correction on the classification of cardiac abnormalities on large heterogeneous datasets of electrocardiogram (ECG) signals. Approach. A state-of-the-art self-learning multi-class label correction method for image classification is adapted to learn a multi-label classifier for electrocardiogram signals. We evaluated our performance using 5-fold cross-validation on the publicly available PhysioNet/Computing in Cardiology (CinC) 2021 Challenge data, with full and reduced sets of leads. Due to the unknown label noise in the testing set, we tested our approach on the MNIST dataset. We investigated the performance under different levels of structured label noise for both datasets. Main results. Under high levels of noise, the cross-validation results of self-learning label correction show an improvement of approximately 3% in the challenge score for the PhysioNet/CinC 2021 Challenge dataset and an improvement in accuracy of 5% and reduction of the expected calibration error of 0.03 for the MNIST dataset. We demonstrate that self-learning label correction can be used to effectively deal with the presence of unknown label noise, also when using a reduced number of ECG leads.

Impact of Staining Methods and Human Factors on Accuracy of Manual Reticulocyte Enumeration.

Although peripheral blood reticulocyte enumeration reflects bone marrow functional integrity, which is important for differential diagnosis of hematological diseases, the factors affecting its accuracy have not been adequately addressed. Using 100 consecutive venous blood samples being processed with four supravital staining techniques [i.e., brilliant cresyl blue (BCB), new methylene blue (NMB), and BCB/NMB with Liu’s stain] for reticulocyte enumeration, two technologists (senior vs. junior) conducted microscopic counting. The results were compared with those obtained with an automated system (Sysmex XE-5000) that served as the standard. The aims of this study were to identify (1) the technique that gave the most reliable outcome, and (2) possible human factors (i.e., seniority, repeated counting) that may affect the counting results. Analysis showed least bias (i.e., deviation from automated counting) associated with BCB staining, followed by NMB. In addition, the senior observer exhibited a higher bias in counting compared with their junior counterpart. Repeated counting also correlated with a higher rate of bias. Nevertheless, inter-observer consistency was high (intraclass correlation coefficient >0.95) and inter-/intra-observer variations were non-significant (both p > 0.05). Our results supported the use of BCB stain for reticulocyte enumeration and the reliability of manual counting despite the involvement of human factors, which had negligible impacts on the final outcomes.

Accuracy of new transthoracic 3D echocardiographic automated algorithm for right ventricle quantification in children with congenital heart disease

Management of many congenital heart diseases (CHD) depends on right ventricle (RV) assessment for which cardiac magnetic resonance (CMR) remains the gold standard. The objective of this study was to assess feasibility and reproducibility of a new semi-automated algorithm to quantify RV (4D auto RVQ) from 3D transthoracic echocardiography (3D TTE) in a pediatric population, and to compare measurements with those obtained from CMR. In all, 90 infants were prospectively included (age 7.5 years, 3–18). Fifteen healthy infants and 75 patients with CHD (30 repaired TOF, 25 ASD, 8 dilated pulmonary valve stenosis, 7 tricuspid regurgitation, 3 APVR, and 2 repaired TAC). All patients underwent 2D and 3D TTE (Vivid E95, GE Healthcare). RV end-diastolic volume (RVEDV), RV end-systolic volume (RVESV), RV ejection volume (RVEV) and RV ejection fraction (RVEF) were analyzed by 4D auto RVQ and compared with measurements from 18 patients who underwent CMR exams for the follow up of their cardiopathy. 4D auto RVQ was feasible in 87 infants (96.7%). In 3 patients 4D auto RVQ was unable to detect RV contouring due to inadequate image quality. Manual contour adjustment was necessary in 81 cases (80%). Mean time of analysis including the time of manual adjustment was 62 seconds. RV was enlarged in 65 of 75 patients with CHD (mean 95 ml/m 2 , 62–165 ml/m 2 ). Good correlation for RVEDV, RVESV, RVEV and RVEF was found between 4D auto RVQ and CMR (r = 0.89–0.99, all P < 0.0001). 4D auto RVQ underestimated RVEDV/BSA of 8 ml/m 2 and RVEF by 6%, and overestimated RVESV/BSA of 9 ml/m 2 . Intraobserver and interobserver variability were low (4.7% and 8.9% respectively) ( Fig. 1 ). 4D auto RVQ is a promising tool for assessment of RV volume and function in children. Such simple and non-invasive exam could reduce the practice of CMR imaging during the follow up of children with CHD.

Leveraging CMR for 3D echocardiography: an annotated multimodality dataset for AI

Abstract Funding Acknowledgements: Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Health Research Council of New Zealand (HRC) National Heart Foundation of New Zealand (NHF) Segmentation of the left ventricular myocardium and cavity in 3D echocardiography (3DE) is a critical task for the quantification of systolic function in heart disease. Continuing advances in 3DE have considerably improved image quality, prompting increased clinical uptake in recent years, particularly for volumetric measurements. Nevertheless, analysis of 3DE remains a difficult problem due to inherently complex noise characteristics, anisotropic image resolution, and regions of acoustic dropout. One of the primary challenges associated with the development of automated methods for 3DE analysis is the requirement of a sufficiently large training dataset. Historically, ground truth annotations have been difficult to obtain due to the high degree of inter- and intra-observer variability associated with manual 3DE segmentation, thus, limiting the scope of AI-based solutions. To address the lack of expert consensus, we instead used labels derived from cardiac magnetic resonance (CMR) images of the same subjects. By spatiotemporally registering CMR labels to corresponding 3DE image data on a per subject basis (Figure 1), we collated 520 annotated 3DE images from a mixed cohort of 130 human subjects (2 independent single-beat acquisitions per subject at end-diastole and end-systole) consisting of healthy controls and patients with acquired cardiac disease. Comprising images acquired across a range of patient demographics, this curated dataset exhibits variation in image quality, 3DE acquisition parameters, as well as left ventricular shape and pose within the 3D image volume. To demonstrate the utility of such a dataset, nn-UNet, a self-configuring deep learning method for semantic segmentation was employed. An 80/20 split of the dataset was used for training and testing, respectively, and data augmentations were applied in the form of scaling, rotation, and reflection. The trained network was capable of reproducing measurements derived from CMR for end-diastolic volume, end-systolic volume, ejection fraction, and mass, while outperforming an expert human observer in terms of accuracy as well as scan-rescan reproducibility (Table I). As part of ongoing efforts to improve the accuracy and efficiency of 3DE analysis, we have leveraged the high resolution and signal-to-noise-ratio of CMR (relative to 3DE), to create a novel, publicly available benchmark dataset for developing and evaluating 3DE labelling methods. This approach not only significantly reduces the effects of observer-specific bias and variability in training data arising from conventional manual 3DE analysis methods, but also improves the agreement between cardiac indices derived from 3DE and CMR. Figure 1. Data annotation workflow Table I. Results

Measuring the margin distance in pulmonary wedge resection.

Margin distance contributes to survival and recurrence during wedge resections for early-stage non-small cell lung cancer. The Initiative for Early Lung Cancer Research on Treatment sought to standardize a surgeon-measured margin intraoperatively. Lung cancer patients who underwent wedge resection were reviewed. Margins were measured by the surgeon twice as per a standardized protocol. Intraobserver variability as well as surgeon-pathologist variability were compared. Forty-five patients underwent wedge resection. Same-surgeon measurement analysis indicated good reliability with a small mean difference and narrow limit of agreement for the two measures. The median surgeon-measured margin was 18.0 mm, median pathologist-measured margin was 16.0 mm and the median difference between the surgeon-pathologist margin was -1.0 mm, ranging from -18.0 to 12.0 mm. Bland-Altman analysis for margin measurements demonstrated a mean difference of 0.65 mm. The limit of agreement for the two approaches were wide, with the difference lying between -16.25 and 14.96 mm. A novel protocol of surgeon-measured margin was evaluated and compared with pathologist-measured margin. High intraobserver agreement for repeat surgeon measurements yet low-to-moderate correlation or directionality between surgeon and pathologic measurements were found. A standardized protocol may reduce variability in pathologic assessment. These findings have critical implications considering the impact of margin distance on outcomes.

Validation of a proposed radiographic bone defect classification system

ObjectivesTo clinically validate a recently proposed classification scheme of post-traumatic bone defects. Methods and materialsOpen fractures were classified utilising a newly introduced classification system. This classification system is based on plain radiographs, assessing the extent and local geometry of bone loss, including: D1 - Incomplete Defects; D2 - Minor/Sub-Critical (Complete) Defects (<2 cm); and D3 - Segmental/Critical Sized Defects (≥2 cm). Reliability was assessed among six independent assessors (three trauma orthopaedic surgeons and three orthopaedic training surgeons) using Fleiss’ kappa tests. 43 open fractures from a tertiary referral trauma centre and their radiographic series were analysed. ResultsInterobserver reliability testing demonstrated the proposed classification system had substantial agreement between the 6 observers, κ = 0.623 (z = 33.8), p < 0.001. Intraobserver variability showed a range of substantial to almost perfect agreement of each observer following a three-week interval between repeat assessments, κ range 0.69–0.914, p < 0.001. ConclusionsIn this representative validation cohort there was substantial agreement between observers when assessing a diverse range of bone defects following long bone open trauma, with highly reproducible assessments by both orthopaedic surgeons and trainee orthopaedic surgeons alike on an internal level. The classification scheme is based on conventional orthogonal radiographs and requires no sophisticated technology, and is therefore pragmatic and applicable to secondary, tertiary and quaternary levels of care for trauma patients. Level of EvidenceIII

Uncertainty-driven ensembles of multi-scale deep architectures for image classification

The use of automatic systems for medical image classification has revolutionized the diagnosis of a high number of diseases. These alternatives, which are usually based on artificial intelligence (AI), provide a helpful tool for clinicians, eliminating the inter and intra-observer variability that the diagnostic process entails. Convolutional Neural Network (CNNs) have proved to be an excellent option for this purpose, demonstrating a large performance in a wide range of contexts. However, it is also extremely important to quantify the reliability of the model’s predictions in order to guarantee the confidence in the classification. In this work, we propose a multi-level ensemble classification system based on a Bayesian Deep Learning approach in order to maximize performance while providing the uncertainty of each classification decision. This tool combines the information extracted from different architectures by weighting their results according to the uncertainty of their predictions. Performance is evaluated in a wide range of real scenarios: in the first one, the aim is to differentiate between different pulmonary pathologies: controls vs bacterial pneumonia vs viral pneumonia. A two-level decision tree is employed to divide the 3-class classification into two binary classifications, yielding an accuracy of 98.19%. In the second context, performance is assessed for the diagnosis of Parkinson’s disease, leading to an accuracy of 95.31%. The reduced preprocessing needed for obtaining this high performance, in addition to the information provided about the reliability of the predictions evidence the applicability of the system to be used as an aid for clinicians.

The "stiff rim" sign of hepatocellular carcinoma on shear wave elastography: correlation with pathological features and potential prognostic value.

To explore the pathologic basis, the influencing factors and potential prognostic value of the stiff rim sign in two-dimensional shear wave elastography (2D-SWE) of hepatocellular carcinoma (HCC). HCC patients who underwent tumor 2D-SWE examination before resection were prospectively enrolled. The stiff rim sign was defined as increased stiffness in the peritumoral region. Interobserver and intraobserver variability of the stiff rim sign was assessed. The correlation between the stiff rim sign and pathological characteristics was analyzed. Multivariate analysis was performed to examine clinical and radiological factors influencing the appearance of stiff rim sign. The Kaplan-Meier method was used to analyze the relationship between recurrence-free survival (RFS) and the stiff rim sign. The stiff rim sign on 2D-SWE was present in 44.7% of HCC lesions. Interobserver agreement and intraobserver agreement for the stiff rim sign were substantial (κ = 0.772) and almost perfect (κ = 0.895), respectively. Pathologically, the stiff rim sign was associated with capsule status, capsule integrity, capsule thickness, proportion of peritumoral fibrous tissue, and peritumoral fibrous arrangement. Multivariate analysis showed that tumor size was an independent clinical predictor for the appearance of stiff rim sign (OR 1.201, p = 0.008). Kaplan-Meier analysis showed RFS was significantly poorer in the stiff rim sign (+) group than the stiff rim sign (-) group in solitary tumors smaller than 5cm (p = 0.007) and solitary tumors with intratumoral stiffness less than 33.7kPa (p = 0.007). The stiff rim sign on 2D-SWE was mainly correlated with peritumoral fibrous tissue status and was a poor prognostic indicator for HCC.

Kinetics of capillary refill time after fluid challenge

BackgroundCapillary refill time (CRT) is a valuable tool for triage and to guide resuscitation. However, little is known about CRT kinetics after fluid infusion.MethodsWe conducted a prospective observational study in a tertiary teaching hospital. First, we analyzed the intra-observer variability of CRT. Next, we monitored fingertip CRT in sepsis patients during volume expansion within the first 24 h of ICU admission. Fingertip CRT was measured every 2 min during 30 min following crystalloid infusion (500 mL over 15 min).ResultsFirst, the accuracy of repetitive fingertip CRT measurements was evaluated on 40 critically ill patients. Reproducibility was excellent, with an intra-class correlation coefficient of 99.5% (CI 95% [99.3, 99.8]). A CRT variation larger than 0.2 s was considered as significant. Next, variations of CRT during volume expansion were evaluated on 29 septic patients; median SOFA score was 7 [5–9], median SAPS II was 57 [45–72], and ICU mortality rate was 24%. Twenty-three patients were responders as defined by a CRT decrease > 0.2 s at 30 min after volume expansion, and 6 were non-responders. Among responders, we observed that fingertip CRT quickly improved with a significant decrease at 6–8 min after start of crystalloid infusion, the maximal improvement being observed after 10–12 min (−0.7 [−0.3;−0.9] s) and maintained at 30 min. CRT variations significantly correlated with baseline CRT measurements (R = 0.39, P = 0.05).ConclusionsCRT quickly improved during volume expansion with a significant decrease 6–8 min after start of fluid infusion and a maximal drop at 10–12 min.

Intraobserver Reliability on Classifying Bursitis on Shoulder Ultrasound.

Purpose: Bursitis is a common musculoskeletal cause of shoulder pain and treatment varies, thus correctly diagnosing and grading bursitis is paramount in deciding management. Our aim was to assess reliability in grading shoulder bursitis on ultrasonography among fellowship trained musculoskeletal radiologists at our institution. Methods: Retrospective study of patients diagnosed with bursitis on ultrasonography. Single-sonographic images of the subacromial-subdeltoid bursa were collected for each patient and randomized to form a test-bank of varying degrees of bursitis. Three months after the test was administered, the cases were randomized and readministered. The radiologists graded each case as: within normal limits, mild, moderate or severe. Intraobserver variability was measured using Cohen's kappa coefficient. Linear regression model was performed to assess correlation between years of experience and kappa. Results: 10 radiologists reviewed 70 cases of bursitis. Kappa values ranged from .53 to .91, indicating 'moderate' to 'almost perfect' variability amongst radiologists. A moderate positive correlation of improving variability (r = .69) with increasing years of experience exists. Conclusion: Fellowship trained musculoskeletal radiologists were able to grade shoulder bursitis with moderate to almost perfect variability, with a positive correlation of improved variability with increasing experience. This may help clinicians choose the correct treatment more confidently in their patients with shoulder pain.

Deep learning-based automatic segmentation for size and volumetric measurement of breast cancer on magnetic resonance imaging

PurposeIn clinical work, accurately measuring the volume and the size of breast cancer is significant to develop a treatment plan. However, it is time-consuming, and inter- and intra-observer variations among radiologists exist. The purpose of this study was to assess the performance of a Res-UNet convolutional neural network based on automatic segmentation for size and volumetric measurement of mass enhancement breast cancer on magnetic resonance imaging (MRI).Materials and methodsA total of 1,000 female breast cancer patients who underwent preoperative 1.5-T dynamic contrast-enhanced MRI prior to treatment were selected from January 2015 to October 2021 and randomly divided into a training cohort (n = 800) and a testing cohort (n = 200). Compared with the masks named ground truth delineated manually by radiologists, the model performance on segmentation was evaluated with dice similarity coefficient (DSC) and intraclass correlation coefficient (ICC). The performance of tumor (T) stage classification was evaluated with accuracy, sensitivity, and specificity.ResultsIn the test cohort, the DSC of automatic segmentation reached 0.89. Excellent concordance (ICC > 0.95) of the maximal and minimal diameter and good concordance (ICC > 0.80) of volumetric measurement were shown between the model and the radiologists. The trained model took approximately 10–15 s to provide automatic segmentation and classified the T stage with an overall accuracy of 0.93, sensitivity of 0.94, 0.94, and 0.75, and specificity of 0.95, 0.92, and 0.99, respectively, in T1, T2, and T3.ConclusionsOur model demonstrated good performance and reliability for automatic segmentation for size and volumetric measurement of breast cancer, which can be time-saving and effective in clinical decision-making.

Evaluation of a Fluoroscopic Angiography Protocol in Hispaniolan Amazon Parrots (Amazona ventralis).

Fluoroscopic angiography evaluates the heart and vascular tree in real time and can be recorded for further diagnostic analysis and measurements. Although reports have been published of the use of fluoroscopic angiography in birds, this technique has not been evaluated in any avian species. The purpose of this study was to evaluate a fluoroscopic angiography protocol in 12 adult Hispaniolan Amazon parrots (Amazona ventralis). Under general anesthesia, the birds were positioned in right lateral (LAT) recumbency on a fluoroscopy table. A bolus of nonionic iodinated contrast agent was injected through a catheter inserted into the basilic or medial metatarsal vein during video acquisition. The same bolus was repeated to obtain the ventrodorsal (VD) view with the bird placed in dorsal recumbency. Eleven studies were performed. A total of 19 (10 VD, 9 LAT) continuous, real-time, fluoroscopic angiograms were successfully captured. The brachiocephalic trunk, aorta, pulmonary arteries, pulmonary veins, and caudal vena cava were visualized, and selected intraluminal measurements collected. The intraobserver and interobserver variability for 3 observers was calculated. Intraobserver agreement was found to be near perfect (intraclass correlation coefficient ≥0.95), whereas interobserver agreement was moderate to substantial (intraclass correlation coefficient ≥0.52). Coefficients of variation were excellent (VD 0.99, LAT 0.99) for intraobserver assessments and moderate (VD 0.72, LAT 0.52) for interobserver assessments. For the interobserver assessments, the VD projection measurements performed better than the LAT measurements. These results suggest that although there was some variation between different observers, relatively consistent vascular measurements could be obtained. The described fluoroscopic angiography protocol is a repeatable and reliable technique that could be useful for the diagnosis and monitoring of cardiovascular diseases in birds.

Comparison between reliability of Schatzker's classification and CT based four quadrant classification for tibial plateau fractures

IntroductionConventionally used Schatzker and AO/OTA classification, do not identify posterior column injuries in tibial plateau fractures. CT based Four quadrant classification, has been proposed to identify fractures of tibial plateau and help in surgical planning of treatment. However, to assess its validity, there is scanty literature about its reliability (inter and intra-observer variation) and comparison of its reliability with that of the more popular Schatzker classification in classifying tibial plateau fracture. Material and methodsX-rays and CT scans of 35 patients (18–65 years) of closed tibial plateau fractures were assessed by 5 Orthopaedic surgeons and classified by both Schatzker's and Four Quadrant Classification. A similar observation was recorded after 2 months. All observers were blinded for the demographic and clinical details of the patients. Their responses were noted and interobserver and Intraobserver variation was calculated. Kappa Test of Cohen was used to determine the level of agreement, as per Landis and Koch's criteria. The reliability of four quadrant classification was also compared with that of Schatzker's classification. ResultsThere was a moderate agreement in interobserver variation in observations for Schatzker's classification (on X-rays) which improved to substantial agreement when the observations were recorded after showing both CT and X-rays. The intraobserver variation had substantial agreement. On the other hand, there was a perfect agreement in both intra- and interobserver variation for Four Quadrant Classification. This difference between the reliability of both classifications was statistically significant (p < 0.001). ConclusionsFour Quadrant classification is a more reliable classification having a better agreement on interobserver and intraobserver variation.