Bland-Altman Research Articles

Reliability of ultrasound measurements of muscle thickness and echointensity in older adults with type 2 diabetes

Ultrasound is an emerging tool for muscle-based measurements. However, the analysis of hyperechoic images (high muscle fat infiltration, as in type 2 diabetes [T2D]) can be challenging. ObjectiveTo evaluate the reliability of ultrasound image analysis of muscle thickness and echointensity in older adults with T2D and non-diabetic controls by experienced and novice analysts. We hypothesized that reliability would be high for the experienced analysts, and lower, but acceptable, for a novice analyst. MethodsWe recruited three groups: 1) adults >60 years with T2D; 2) age- and sex-matched normoglycemic controls; and 3) healthy adults 18–35 years. All participants underwent ultrasound imaging of the abdomen and anterior thigh. Ultrasound images were de-identified, randomized, and separately analyzed by two expert analysts (>3 years experience each). A subset of images was also analyzed by a novice analyst with no prior experience. Inter- and intra-rater reliability was assessed using coefficients of variation (CVs), intraclass correlation coefficients (ICCs), and Bland-Altman analysis. ResultsFor the expert analysts, ICCs were >0.90 and CVs were <10 % for all measurements, regardless of participant group. For the novice analyst, all ICCs were >0.90, except for rectus abdominus thickness measurements in the T2D group (ICC: 0.659 [95%CI: -0.134, 0.911]). CVs were <10 % for all measurements, except for rectus abdominus thickness in the T2D group (CV: 12.7 %). Ultrasound image analysis of muscle thickness and echointensity by experienced analysts was highly reliable. ConclusionNovice and expert analysts produced comparable measurements, except for rectus abdominus thickness in images obtained from older adults with T2D.

A Retrospective Database Analysis to Investigate if Electrolytes in Venous Blood are Equivalent to the Levels in Arterial Blood.

In a critically ill patient, when an arterial blood sample is processed on an arterial blood gas (ABG) analyzer, it also measures electrolytes apart from analyzing the blood gases. The turnaround time for ABG analysis is way too less compared to the conventional electrolyte analysis with a serum sample. This study intends to investigate whether values of electrolytes estimated in arterial blood can substitute the routinely practiced method. This is a retrospective cross-sectional study. The source of data is patients' reports of serum electrolytes and ABG analysis from the Clinical Biochemistry laboratory, CIMS Teaching Hospital, Chamarajanagar between January and June 2021. The electrolytes report of 200 patients from whom both arterial and venous blood samples were sent to the Clinical Biochemistry laboratory on the same day and at the same time for analysis were selected. The data was compiled, compared, and correlated using a suitable statistical tool. The mean and standard deviation of sodium (135.62 ± 5.20 in venous vs 134.08 ± 8.49 in arterial blood), potassium (4.20 ± 0.64 vs 3.80 ± 0.75), and chloride (102.28 ± 4.99 vs 96.33 ± 8.11) were observed. However, when the concordance correlation coefficient and Bland-Altman plot analysis were made there was no agreement between electrolytes analyzed on serum in an autoanalyzer with that of ABG analyzer. We conclude that the electrolytes measured by a conventional autoanalyzer on a serum sample cannot be replaced by values analyzed on a blood gas analyzer. Devaki RN, Kasargod P, Roopa Urs AN, Chandrika N. A Retrospective Database Analysis to Investigate if Electrolytes in Venous Blood are Equivalent to the Levels in Arterial Blood. Indian J Crit Care Med 2024;28(5):442-446.

Anterior segment swept-source optical coherence tomography and ultrasound biomicroscopy in iris and ciliary body lesions

ABSTRACT Objectives Differentiation of iris and ciliary body lesions as benign or malignant and cystic or solid is important. The aim of this study was to compare anterior segment swept-source optical coherence tomography (AS SS-OCT) and ultrasound biomicroscopy (UBM) findings in iris and ciliary body tumors. Research design and methods Forty-two eyes of 38 cases with iris and ciliary body tumors imaged with UBM and AS SS-OCT between September 2018 and September 2023 were evaluated retrospectively. Results Of 42 eyes, 14 had melanoma, 14 iris pigment epithelial (IPE) cysts, 7 nevi, 3 Lisch nodules, 2 iris stromal cysts, 1 pars plana cysts, and 1 iris mammillations. An equivalent (100%) visualization of the anterior tumor margin was obtained with both techniques. Compared to AS SS-OCT, UBM was superior for posterior margin visualization in melanocytic tumors and IPE cysts. Bland-Altman plots demonstrated good agreement between UBM and AS SS-OCT for melanocytic tumors < 2.5 mm in base diameter and < 2 mm in thickness. Conclusions Although, UBM is the gold standard for ciliary body and iridociliary tumors. AS SS-OCT should be considered as an excellent alternative to UBM, especially in minimally elevated iris lesions.

Neonatal Endotracheal Intubation: Comparison of Existing Formulae for Accuracy

Background There are several formulae proposed to calculate the ideal depth of insertion of the endotracheal tube (ETT) in neonates, none of which are universally accurate. Our objective is to identify the most accurate of these formulae and to determine the most suitable formula for babies weighing less than 1 kg. Methods This is a retro-prospective observational study conducted in a tertiary neonatal intensive care unit. All intubated babies, irrespective of their gestational age, were included in this study. Those with congenital airway anomalies or syndromes with involvement of the upper airway and those who required short-term ventilation restricted to the theatre complex were excluded. The ideal insertion depth was determined via a chest X-ray (CXR) with the ETT projecting between thoracic vertebrae 2 and 3. Predictions from eight formulae investigated in the Indian population were analysed using paired Wilcoxon tests and Bland–Altman plots. Results The sample size was 83, of which 16 weighed less than 1 kg. Neither weight-based nor naso-tragal length (NTL)-based formulae were universally accurate in the prediction of the ideal depth of insertion, and using either of the formulae resulted in a high rate of malpositioning. Amongst the NTL-based formulae, Tatwavedi and Singh’s adjustments worked better than the original. Conclusion The currently available formulae remain inaccurate in the prediction of the ideal length of insertion of ETT, more so in babies less than 1 kg in our study population.

CycleGAN-Driven MR-Based Pseudo-CT Synthesis for Knee Imaging Studies

In the field of knee imaging, the incorporation of MR-based pseudo-CT synthesis holds the potential to mitigate the need for separate CT scans, simplifying workflows, enhancing patient comfort, and reducing radiation exposure. In this work, we present a novel DL framework, grounded in the development of the Cycle-Consistent Generative Adversarial Network (CycleGAN) method, tailored specifically for the synthesis of pseudo-CT images in knee imaging to surmount the limitations of current methods. Upon visually examining the outcomes, it is evident that the synthesized pseudo-CTs show an excellent quality and high robustness. Despite the limited dataset employed, the method is able to capture the particularities of the bone contours in the resulting image. The experimental Mean Absolute Error (MAE), Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), Zero-Normalized Cross Correlation (ZNCC), Mutual Information (MI), Relative Change (RC), and absolute Relative Change (|RC|) report values of 30.4638 ± 7.4770, 28.1168 ± 1.5245, 0.9230 ± 0.0217, 0.9807 ± 0.0071, 0.8548 ± 0.1019, 0.0055 ± 0.0265, and 0.0302 ± 0.0218 (median ± median absolute deviation), respectively. The voxel-by-voxel correlation plot shows an excellent correlation between pseudo-CT and ground-truth CT Hounsfield units (m = 0.9785; adjusted R2 = 0.9988; ρ = 0.9849; p &lt; 0.001). The Bland–Altman plot shows that the average of the differences is low ((HUCT−HUpseudo−CT = 0.7199 ± 35.2490; 95% confidence interval [−68.3681, 69.8079]). This study represents the first reported effort in the field of MR-based knee pseudo-CT synthesis, shedding light to significantly advance the field of knee imaging.

Quantification of infarct core volume in patients with acute ischemic stroke using cerebral metabolic rate of oxygen (CMRO2) in CT perfusion.

The cerebral metabolic rate of oxygen (CMRO2) is considered a robust marker of the infarct core in 15°-tracer- based positron emission tomography. We aimed to delineate the infarct core in patients with acute ischemic stroke using commonly used relative cerebral blood flow (rCBF) < 30% and oxygen metabolism parameter of CMRO2 on CT perfusion in comparison with pre-treatment diffusion- weighted imaging (DWI)-derived infarct core volume. Patients with acute ischemic stroke who met the inclusion criteria were recruited. The CMRO2 and CBF maps in CT perfusion were automatically generated using post-processing software. The infarct core volume was quantified with relative (r) CMRO2 < 20% - 30% and rCBF < 30%. The optimal threshold was defined as those that demonstrated the smallest mean absolute error, lowest mean infarct core volume difference, narrowest 95% limit of agreement, and largest intraclass correlation coefficient (ICC) against the DWI. This study included 76 patients (mean age ± standard deviation, 69.97 ± 12.15 years, 43 males). The optimal thresholds of rCMRO2 < 26% resulted in the lowest mean infarct core volume difference, narrowest 95% limit of agreement, and largest ICC among different thresholds. Bland-Altman analysis demonstrated a volumetric bias of 1.96 mL between DWI and rCMRO2 < 26%, whereas in cases of DWI and rCBF < 30%, the bias was notably larger at 14.10 mL. The highest correlation was observed for rCMRO2 < 26% (ICC=0.936), whereas rCBF < 30% showed a slightly lower ICC of 0.934. CT perfusion-derived CMRO2 is a promising parameter for estimating the infarct core volume in patients with acute ischemic stroke. CMRO2 = cerebral metabolic rate of oxygen.

Artificial intelligence model system for bone age assessment of preschool children.

Our study aimed to assess the impact of inter- and intra-observer variations when utilizing an artificial intelligence (AI) system for bone age assessment (BAA) of preschool children. A retrospective study was conducted involving a total sample of 53 female individuals and 41 male individuals aged 3-6 years in China. Radiographs were assessed by four mid-level radiology reviewers using the TW3 and RUS-CHN methods. Bone age (BA) was analyzed in two separate situations, with/without the assistance of AI. Following a 4-week wash-out period, radiographs were reevaluated in the same manner. Accuracy metrics, the correlation coefficient (ICC)and Bland-Altman plots were employed. The accuracy of BAA by the reviewers was significantly improved with AI. The results of RMSE and MAE decreased in both methods (p < 0.001). When comparing inter-observer agreement in both methods and intra-observer reproducibility in two interpretations, the ICC results were improved with AI. The ICC values increased in both two interpretations for both methods and exceeded 0.99 with AI. In the assessment of BA for preschool children, AI was found to be capable of reducing inter-observer variability and enhancing intra-observer reproducibility, which can be considered an important tool for clinical work by radiologists. The RUS-CHN method is a special bone age method devised to be suitable for Chinese children. The preschool stage is a critical phase for children, marked by a high degree of variability that renders BA prediction challenging. The accuracy of BAA by the reviewers can be significantly improved with the aid of an AI model system. This study is the first to assess the impact of inter- and intra-observer variations when utilizing an AI model system for BAA of preschool children using both the TW3 and RUS-CHN methods.

Correlation between the radial artery resistance index and the systemic vascular resistance index: a cross-sectional study.

Ultrasound measurement of the radial resistance index (RRI) in the anatomical snuffbox has been proposed as a useful method for assessing the systemic vascular resistance index (SVRI). This study aims to establish the correlation between SVRI measured by pulmonary artery catheter (PAC) and RRI. A cross-sectional study included all consecutive patients undergoing postoperative (POP) cardiac surgery with hemodynamic monitoring using PAC. Hemodynamic assessment was performed using PAC, and RRI was measured with ultrasound in the anatomical snuffbox. The Pearson correlation test was used to establish the correlation between RRI and SVRI measured using PAC. Hemodynamic behavior concerning RRI with a cutoff point of 1.1 (described to estimate under SVRI) was examined. Additionally, consistency between two evaluators was assessed for RRI using the intraclass correlation coefficient and Bland-Altman analysis. A total of 35 measurements were obtained. The average cardiac index (CI) was 2.73 ± 0.64L/min/m², and the average SVRI was 1967.47 ± 478.33 dyn·s·m²/cm5. The correlation between RRI and SVRI measured using PAC was 0.37 [95% CI 0.045-0.62]. The average RRI was 0.94 ± 0.11. RRI measurements > 1.1 had a mean SVRI of 2120.79 ± 673.48 dyn·s·m²/cm5, while RRI measurements ≤ 1.1 had a mean SVRI of 1953.1 ± 468.17 dyn·s·m²/cm5 (p = 0.62). The consistency between evaluators showed an intraclass correlation coefficient of 0.88 [95% CI 0.78-0.93], and Bland-Altman analysis illustrated adequate agreement of RRI evaluators. For patients in cardiac surgery POP, the correlation between the SVRI measured using PAC and the RRI measured in the anatomical snuffbox is low. Using the RRI as a SVRI estimator for patients is not recommended in this clinical scenario.

Biological variation of 16 biochemical analytes estimated from a large clinicopathologic database of dogs and cats.

Biochemical measurements are commonly evaluated using population-based reference intervals; however, there is a growing trend toward reassessing results with within-subject variation (CVI). We aimed to estimate the CVI of 16 biochemical analytes using a large database of dogs and cats, which refers to the results of routine health checkups. Pairs of sequential results for 16 analytes were extracted from a database of adult patients. The second result was divided by the first result to produce the ratio of sequential results (rr), and the frequency distribution of rr was plotted. From the plots, the coefficient of variation (CVrr) was calculated. Analytical variation (CVA) was calculated using quality control data, and CVI was estimated as follows: . Estimated CVI was compared with previously reported CVI using the Bland-Altman plot analysis. From the database, 9078 data points from 3610 dogs and 3743 data points from 1473 cats were extracted, with 5468 data pairs for dogs and 2270 for cats. Sampling intervals ranged from 10 to 1970 days (median 366) for dogs and 23 to 1862 days (median 365) for cats. Bland-Altman analysis showed most CVI plots fell within the limits of agreement; however, positive fixed biases were observed in both dogs and cats. Our study introduces a novel approach of estimating CVI using routine health checkup data in dogs and cats. Despite biases, our method holds promise for clinical application in assessing the significance of measurement result differences.

To Determine the Accuracy and Reliability of Broselow Tape in Comparison with Actual Weight and Advance Pediatric Life Support Formula Weight Assessment among the Pediatric Population Presenting to Emergency Department of a Low Income Setting

Background: Near precise weight assessment among sick pediatric patients remains a dilemma and conventionally accepted weight assessment methods, in busy ED might be inaccurate or unreliable. Children have different weights at different ages, and accurate/precise weight measurement is of utmost importance for weight-related dose calculation of drugs/fluids, equipment sizes, an accurate dose of electrical currents during cardiac shock, etc. Several weight estimation methods are available and are dependent on the child's age, length, or both. However, length or length-and-age-based methods may have greater accuracy than merely age-based, still precise weight measurement while children are recumbent (length) has its own challenges. Objective: We determined the accuracy and reliability of BT by comparing it with actual weight and advanced pediatric life support formula (APLS) among the pediatric population presenting to the emergency department. Methods: This was a single-center, cross-sectional study design. This study was conducted at pediatric emergency of an urban tertiary care hospital after ethics committee approval and written consent from parents/caregivers during July 2021- June 22. Pediatric patients aged 1 month to 12 years, weight 3-36 kg, and height 46.5-142.5 cm on BT were included. Actual weight was measured on a standard weighing machine. We use Broselow pediatric emergency tape (2017 edition), APLS formula was also used to measure the estimated weight by using the age provided by parents. Descriptive analysis, mean and standard deviation were calculated, frequencies and percentages were calculated for categorical variables. Cronbach’s alpha and Passing-Bablok regression analysis was applied to assess the reliability and identify systematic biases between actual body weights with estimated BT. Bland–Altman analysis was also performed to measure the precision, accuracy, and bias. Results: 250 children were included with equal gender distribution and were divided in to three categories as per the weight estimation by BT in to &lt;10 kg (n=58, 23.2%), 10-18 kg (n=151, 60.4%) and &gt;18 kg (n=41, 16.4%). The mean age was 5.26 (±2.37) years, majority of children were below 5 years of age (n=144, 58%). Positive agreement between BT weight with actual weight and other formulas in weight category of &lt;10 kg, however as weight increases from 10 kg, onward results are not significant. Passing and Bablok Regression analysis showed a positive correlation between the estimated and actual (AW) body weight (r=0.9280, p&lt;0.001) and accuracy (r2=0.929), and the accuracy of BT weight decreases with the increasing weight of children. Similarly, 95% agreement limit and mean biased was 0.465 to 1.113 and 0.789±2.602 between BT and APLS, BT with AW was -0.50 to -0.28 and -0.39±0.885. Comparing and correlating weight assessment of APLS formula vs LF and TF didn't show significance with a p-value of 0.041 and 0.034 respectively. Bland-Altman plot between BT measurement with AW demonstrates a bias equal to 1.096 kg with a limit of 0.870 to 1.815. Conclusion: BT may be an accurate and time bound method of weight measurement as compared to other methods of weight estimation, however the accuracy of BT may be adversely affected with age exceeding 95 months and weight &gt;26 kg. BT may be safely used in the younger pediatric population. Health care professionals may consider this information while using BT to estimate weight for pediatric resuscitation, however, this is a single center hospital based study with limited cases, we recommend to have a large scale community based study or a census before to generalize these results to general population.

Comparison of infrared thermography of the blowhole mucosa with rectal temperatures in killer whales (Orcinus orca)

Killer whales are an important sentinel species and developing non-invasive methods of health assessments might provide insight for understanding how wildlife health is influenced by ecosystem change. Rectal temperature (RT) is a proxy for core body temperature in managed-care cetaceans, however, this measurement is impractical for free-ranging cetaceans and infrared imaging has been suggested as an alternative. The aim of the current study was to prospectively compare infrared thermography of the blowhole to rectal temperatures in killer whales, as well as establish a healthy range for rectal temperature using retrospective data. Infrared video was recorded from the blowhole of thirteen healthy killer whales in managed care, immediately followed by rectal temperature measurement. Repeated measures Bland-Altman analysis revealed blowhole temperature (BHT) had a bias of -1.28°C from RT. Considerable proportional bias was observed with agreement between measurements improving as mean temperature increased. RT positively associated with air temperature, and inversely associated with body mass. BHT was not significantly affected by sex or body mass but was significantly affected by water temperature and air temperature. Retrospective analysis from eighteen killer whales (n = 3591 observations) was performed to generate expected RT ranges, partitioning out for sex and body mass. Given the proportional bias observed with Bland Altman analysis, BHT cannot currently be recommended as a measurement for absolute core body temperature, however infrared thermography of the blowhole remains a promising tool for health assessment of free-ranging killer whale populations, as it may serve as a non-contact screening tool to detect pyrexic animals within a group.

Comparison of QT interval readings between smartwatch ECG combined with artificial intelligence and 12-lead ECG in subjects hospitalized for antiarrhythmic drug initiation and follow-up

Abstract Background/Introduction Class III antiarrhythmic drugs (AADs) are effective for the maintenance of sinus rhythm (SR) in subjects with atrial fibrillation (AF). However, several require initial hospitalization for QT interval monitoring using the 12-lead ECG. Purpose The purpose of this study was to evaluate the feasibility of using a smartwatch single lead ECG combined with a deep learning algorithm for routine QT interval determination in the context of AAD initiation. Methods A deep learning model was developed using smartwatch and Holter ECG recordings to estimate corrected QT intervals. To evaluate the performance of this model, we carried out a prospective study on subjects hospitalized for initiation of sotalol and/or dofetilide. Subjects recorded near simultaneous (maximal 2 minutes apart) smartwatch ECGs (Apple Watch Series 6) and 12-lead ECG (Mac 5500 HD ECG or GE Carescape B450) right before and 2-3 hours after drug administration twice a day for 2-3 days. Blinded double annotation followed by adjudication of disagreement provided a reference heart rhythm (NSR or AF) and QTc interval. The tangent method was used for QTc interval determination. Bland-Altman analysis was performed to assess the agreement between QTc measures. Results Fifty subjects (mean age 67.4±10.9 years, 64% male) were enrolled in the study from February 2022 to October 2022. Out of 386 monitoring visits, 321 near simultaneous pairs of ECGs were collected with a mean delay of +15.0±41.1s for the smartwatch ECG compared to the 12-lead ECG. Overall, a mean QTcF reference value of 419.9±34.0 ms was observed (435.2±38.4 ms for QTcB, r=0.83 between annotators[BL2] ). Among the 139 pairs with 12-lead ECG found in NSR, we observed a mean QTcF difference of +19.7 ms and limits of agreements equal to -21.5 ms and 60.8 ms, with AI-facilitated measurement. In comparison, a mean QTcF difference of +22.5 ms and limits of agreements equal to -23.8 ms and 68.8 ms were observed for manual readings of smartwatch ECGs instead. Conclusion(s): Our findings suggest that a smartwatch ECG deep learning approach can estimate the QT interval with a fair agreement compared to standard 12-lead monitoring during AAD initiation. This measurement method appears to be at least as accurate as manual measurements on smartwatch ECGs. With further validation, the smartwatch-based automatic detection of QT interval prolongation, could be used for a faster, safe, and extended outpatient QT interval monitoring.Bland-Altman plots

Comparative analysis of the quantitative point-of-care CareSTART™ total bilirubin with central laboratory total bilirubin assays in neonatal blood samples.

This study aimed to assess and compare the accuracy of point-of-care CareSTART™ S1 Total Bilirubin test with a central laboratory total bilirubin assay using neonatal samples. This study was conducted using 152 paired measurements obtained from 122 neonates admitted to the neonatal intensive care unit. Total serum bilirubin (TSB) levels assayed with the central laboratory assay, laboratory bilirubinomter, trancutaneous bliribubin (TcB) instrument and CareSTART were compared using Bland-Altman analysis. The mean difference between the CareSTART and TSB values was -1.43 mg/dL and the 95% limit of agreement (LoA) was -4.25 to 1.39 mg/dL. CareSTART tended to underestimate total bilirubin concentrations compared with TcB, however, the LoA was narrower due to the smaller SD of mean difference for CareSTART. The CareSTART Total Bilirubin test provides an accurate alternative to TcB for total serum bilirubin measurement. Given its low-cost, ease-of-use, and portability, the use of CareSTART is expected to provide point-of-care measurements, especially in low-resource settings.

Real-world validation of smartphone-based photoplethysmography for heart rate monitoring in atrial fibrillation

Abstract Introduction Photoplethysmography (PPG) based smartphone applications facilitate heart rate monitoring that may enable physicians to guide rate control in patients with atrial fibrillation (AF). Despite extensive validation of PPG-derived heart rate, validation studies in the presence of AF are lacking, particularly in unsupervised free-living conditions. Purpose To evaluate the accuracy of PPG-derived heart rate using a smartphone application in patients with an established diagnosis of AF, both during sinus rhythm and AF, in a real-world setting. Methods Consecutive patients undergoing AF ablation were enrolled four weeks prior to the procedure. They were instructed to assess their heart rate twice daily during eight weeks with a one-minute finger-over-the-camera smartphone-PPG recording. The reference heart rate was obtained from two 30-seconds handheld single-lead electrocardiogram (ECG) recordings performed immediately before and after the PPG recording. Measurements were valid if the heart rate did not change by more than 10% between the first and second ECG and maintained sinus rhythm or AF consistently. Results In total, fifty patients (age 63 ±11 years, 68% male) performed 3407 high-quality PPG measurements accompanied with two high-quality ECGs measurements both in sinus rhythm or AF. After excluding 223 measurements due to inconsistent heart rate between the first and second ECG, 3184 measurements remained. The mean heart rate (71.5 ± 17.3 beats per minute (bpm)) was significantly higher during AF (94.7 ± 19.4 bpm) compared to sinus rhythm (65.7 ± 10.7 bpm) (p-value &amp;lt; 0.001). The heart rate assessment was accurate within a 10% margin in 93.8% of PPG measurements. Inaccurate heart rate assessments were more frequent in AF (161/637; 25.3%) compared to sinus rhythm (35/2547; 1.4%) (p-value &amp;lt; 0.001). PPG underestimated the heart rate in sinus rhythm with 0.4 bpm (95% CI: -0.3 to -0.5 bpm) and in AF with 6.6 bpm (95% CI: -5.8 to -7.4 bpm). The deviation of PPG-derived heart rate from the ECG-derived true heart rate was significantly greater during AF (root mean square error (RMSE) = 11.8 bpm) compared to sinus rhythm (RMSE = 2.2 bpm) (p-value &amp;lt; 0.001). The heart rate correlation plots and corresponding Bland-Altman plots demonstrate a skewed underestimation of the heart rate in AF with PPG, particularly for high heart rates. Conclusions Heart rate can accurately be derived from PPG recordings using a smartphone application in real-world conditions. However, PPG measurements may underestimate high heart rates in the presence of AF.Heart Rate in Sinus RhythmHeart Rate in Atrial Fibrillation

Clinical validation of an artificial intelligence algorithm offering automated corrected QT-interval measurements from a single lead electrocardiogram

Abstract Introduction A novel smartwatch (Withings Scanwatch, SW) offers automated analysis of the QTc based on single-lead ECG recordings. Prior clinical validation studies showed insufficient accuracy when compared to manual measurements based on a 12-lead ECG. Purpose We hypothesized that a novel AI-based algorithm (PulseAI) is capable of measuring QTc-interval based on PDF-exports from a single-lead ECG with improved accuracy compared to the manufacturer’s automated QTc measurements. Methods In this prospective, observational study, consecutive patients underwent a nearly simultaneously 12-lead ECG and a single-lead ECG. Two blinded, independent cardiologists manually interpreted the QT-interval in leads I and II of the 12-lead ECG using the tangent-method, which served as the gold standard. QTc was calculated using Bazett’s formula. QTc measurements were compared using the Bland-Altman method. Results A total of 317 patients (48% female, mean age 63 ± 17 years) were enrolled. QTc-intervals could be automatically calculated by the SW and by the novel AI algorithm in 175 patients (55%). Diagnostic accuracy of SW-ECG for detection of QTc-intervals ≥ 450 ms as quantified by the area under the curve was 0.85 for the manufacturer and 0.9 for novel AI algorithm. In 21 patients (12%) QTc was ≥450ms. The Bland-Altman analysis resulted in a bias of 21 ms [95% limit of agreement (LoA) -48 to 89 ms] for the manufacturer algorithm and 2 ms [95% LoA -35 to 39 ms] for the novel AI algorithm, respectively, when compared with manual QTc-measurement based on the 12-lead ECG. Conclusion A novel AI algorithm outperformed the manufacturer algorithm for automated QTc-measurements based on PDF-exported single-lead ECGs.Central Illustration

Fully automatic quantification for hand synovitis in rheumatoid arthritis using pixel-classification-based segmentation network in DCE-MRI.

A classification-based segmentation method is proposed to quantify synovium in rheumatoid arthritis (RA) patients using a deep learning (DL) method based on time-intensity curve (TIC) analysis in dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). This retrospective study analyzed a hand MR dataset of 28 RA patients (six males, mean age 53.7years). A researcher, under expert guidance, used in-house software to delineate regions of interest (ROIs) for hand muscles, bones, and synovitis, generating a dataset with 27,255 pixels with corresponding TICs (muscle: 11,413, bone: 8502, synovitis: 7340). One experienced musculoskeletal radiologist performed ground truth segmentation of enhanced pannus in the joint bounding box on the 10th DCE phase, or around 5min after contrast injection. Data preprocessing included median filtering for noise reduction, phase-only correlation algorithm for motion correction, and contrast-limited adaptive histogram equalization for improved image contrast and noise suppression. TIC intensity values were normalized using zero-mean normalization. A DL model with dilated causal convolution and SELU activation function was developed for enhanced pannus segmentation, tested using leave-one-out cross-validation. 407 joint bounding boxes were manually segmented, with 129 synovitis masks. On the pixel-based level, the DL model achieved sensitivity of 85%, specificity of 98%, accuracy of 99% and precision of 84% for enhanced pannus segmentation, with a mean Dice score of 0.73. The false-positive rate for predicting cases without synovitis was 0.8%. DL-measured enhanced pannus volume strongly correlated with ground truth at both pixel-based (r = 0.87, p < 0.001) and patient-based levels (r = 0.84, p < 0.001). Bland-Altman analysis showed the mean difference for hand joints at the pixel-based and patient-based levels were -9.46 mm3 and -50.87 mm3, respectively. Our DL-based DCE-MRI TIC shape analysis has the potential for automatic segmentation and quantification of enhanced synovium in the hands of RA patients.

#466 Comparison of CKDEPI vs EKFC equations in estimating glomerular filtration rate in kidney transplant recipients

Abstract Background and Aims In 2020, European Kidney Function Consortium (EKFC) proposed a new equation that estimates glomerular filtration rate (GFR) in European population. Using creatinine and age, as well as CKDEPI, EKFC gives an estimation which has been demonstrated to correlate better to measured GFR, particularly at extreme ages. CKDEPI equation is the most common equation used in our clinical practice. Few studies have compared EKFC with CKDEPI, particularly in kidney transplant (KT) recipients. Our objective is to compare estimation of GFR with EKFC and CKDEPI in KT patients. Method We conduct a retrospective observational study including all consecutive KT performed at our centre from 1993 to 2022 with survival greater than one year, and laboratory tests at third month and one-year post-transplant. Using Creatinine and Age we calculate estimated GFR (eGFR) by CKDEPI and EKFC equations. We compare both results in terms of correlation and agreement in the global sample and classifying by subgroups according to a) Sex, b) Age by 10-year sections, c) Body Mass Index (BMI) as WHO classification and d) presence or absence of Diabetes Mellitus (DM). We use SPSS as statistics program. Results We include 968 KT, depicted in Table 1. CKDEPI and EKFC at 3rd month and EKFC at one-year have a non-normal distribution (Kolmogorov-Smirnov test with p = 0.003, p = 0.035 and p = 0.002, respectively), contrary to normality in CKDEPI at one year (p = 0.058). Levene test for homogeneity of variances shows no differences (3rd month p = 0.102, one year p = 0.165), as well as in subgroups. We perform, besides Mann-Whitney, Student's T-test despite limitations over means without significant differences in any subgroup. Mean global difference at 3rd month is 0.23 ml/min greater by CKDEPI. There are no relevant differences between subgroups related to DM and BMI. We can describe a difference between women (0.85 ml/min) and men (0.05 ml/min), and it is larger in 20-30 years group (3.4 ml/min) and in 70-80 years (2.74 ml/min). As measure of agreement, Intraclass Correlation Coefficient (ICC) shows excellent correlation in all subdivisions (Cronbach's Alpha statistic (CA) &amp;gt; 0.99), without significant decrease in any subgroup; minimum ICC = 0.987 is in 20-30 years group, and maximum ICC = 0.999 in 40-60 years. Bland-Altman plot is shown in Fig. 1, where we appreciate a negative bias in low eGFR against a positive in higher values. Mean global difference at one-year eGFR is 0.25 ml/min greater by CKDEPI. We observe a difference between women (0.75 ml/min) and men (0.04 ml/min), larger in 20-30 years group (2.94 ml/min) and 70-80 years (2.27 ml/min) without significance. ICC shows excellent correlation without significant decrease in any subgroup; minimum ICC = 0.992 is in 20-30 years group, and maximum ICC = 0.999 in 50-60 years. Conclusion We find no significant differences between means in eGFR by CKDEPI and EKFC. We observe a tendency to a greater estimation in women than in men. Both CKDEPI and EKFC equations for eGFR show an excellent concordance in our cohort of KT but differences in scatter analysis in Bland-Altman plot, greater in higher values of eGFR. ICC and CA are both slightly minor in the extreme age groups of our cohort. It matches what described in major studies; these differences may be not clinically relevant in this age range group but are probably greater in patients beyond this range.

#2812 Continuous blood pressure monitoring with a finger cuff is unreliable in chronic hemodialysis patients

Abstract Background and Aims Intradialytic hypotension (IDH) is the most common complication of hemodialysis (HD), which is often related to ultrafiltration. Due to the high prevalence of cardiovascular disease (CVD), compensatory mechanisms often fail. In clinical practice, BP is measured 2x/hour, while additional readings are only performed when symptoms arise. However, since IDH defined by a nadir systolic BP (SBP)&amp;lt;90 mmHg or &amp;lt;100 mmHg (depending on pre-dialysis SBP), regardless of symptoms, is most strongly associated with mortality, current BP measurement frequency may severely underestimate the true incidence of IDH. Continuous BP-monitoring may offer a solution to this issue. The Clearsight device (CSD) is a non-invasive continuous BP-monitoring system that measures arterial pressure and waveforms through a finger cuff by photoplethysmography. Yet, a compromised peripheral circulation may hamper its reliability in HD patients. This study aims to assess the agreement between intradialytic CSD-monitoring and standard oscillometric brachial arterial pressure (oBAP) assessment. Methods 40 Chronic HD patients were cross-over randomized to 4 modalities, each for 2 weeks: [standard HD (S-HD), cool HD (C-HD), low volume hemodiafiltration (LV-HDF) and high-volume HDF (HV-HDF)]. During one session of each second week, BP and heart rate (HR) were continuously recorded with the CSD (BMEYE/Edwards Lifesciences) and every 15 minutes with a brachial arterial cuff (contralateral of the fistula or graft, if applicable). Simultaneous measurements of SBP, diastolic BP (DBP) and HR were compared with descriptive statistics (mean difference [bias, BAP minus CSD], standard deviation (SD) and range) and Bland-Altman analysis. Bias was defined as the mean difference between oBAP and CSD, and 95% limits of agreement (LA) as the mean difference ±1.96 SD. Post-hoc, patients were stratified into tertiles of unexplained missing CSD-readings, to explore whether specific patient characteristics explain failed readings. Results 75% of patients was male, with a mean age 69.7 (±13.5 years) and median dialysis vintage 3.0 years (IQR 1.0-5.8). Diabetes mellitus and CVD were present in 48% and 73%, resp. Of the 2720 paired BP-measurements, 42% CSD and 6% of oBAP-readings were missing, leaving 36% of absent CSD-readings unexplained. Bland-Altman analysis, comparing oBAP and CSD-measurements, revealed a mean bias ± LA of 13.3 ± 44.5 mmHg for SBP, a mean bias ± LA of 13.3 ± 25.5 for DBP, and a mean bias ± LA of −0.03 ± 22.9 for HR. Surprisingly, the group with a high (&amp;gt;38%) proportion of unexplained missing CSD-values was significantly younger compared to the low (&amp;lt;21%) and intermediate group (21-38%) (p = 0.015), whereas no significant differences were observed with respect to dialysis vintage, diabetes mellitus or CVD. Conclusions 1) Compared to the oBAP method, finger cuff-monitoring shows poor precision and accuracy in BP-measurements, but good accuracy and poor precision in HR-readings; 2) 1/3 of readings failed for unknown reasons. Thus, CSD-monitoring is unsuitable for intradialytic hemodynamic monitoring in chronic HD patients.

#2124 Accuracy and interferences in the T50 test: assessing the reliability of a novel diagnostic tool

Abstract Background and Aims A novel blood-based in vitro diagnostic test determines the calcification propensity in patients, e.g. chronic kidney disease (CKD) patients. This assay is known as the ‘T50 test’ and is based on the one-half maximal transition time (T50; minutes) of calciprotein particle 1 (CPP1) to calciprotein particle 2 (CPP2) in serum (Fig. 1). Numerous studies have associated a lower T50 with cardiovascular and all-cause mortality in chronic kidney disease (CKD) patients at stages 3, 4 and 5, kidney transplant recipients, and in the general population. To date, the T50 test has been performed at a single site (Calciscon AG, Biel, Switzerland). The aim of this study was to determine the accuracy of the T50 test. Additionally, the effect of haemolysis, hyperbilirubinemia and lipidemia interferences on the T50 was studied and compared to the manufacturer's claims. Method The T50 was measured using T50 test kits and according to the protocol of Calciscon AG using an automatic pipetting machine (EVOware Freedom, Tecan) and a nephelometer (NEPHELOstar, BMG). Patient serum samples were spiked with haemoglobin, bilirubin, or lipids (triglyceride) for the interference analysis. In total, three dilution series with interferences and four additional non-spiked patient serum samples were included (n = 32). The T50 of all samples were measured at two laboratory sites: Amsterdam UMC (Amsterdam, The Netherlands) and Calciscon AG. Data were statistically analysed for accuracy using the Passing Bablok regression and the Bland Altman plot. Interference was established with deviations &amp;gt;10% in T50 compared to the control sample. Results The measurements of all samples at both sites were similar with a mean bias of 0.24% (±4.84%) and a great agreement was found (slope = 1.012; 95% CI [0.931 to 1.107]) (Table 1). Interference was observed at higher levels of haemoglobin and triglyceride with concentrations of at least 0.1 g/dL and 1124.8 mg/dL, respectively. The triglyceride interference is in agreement with the manufacturer's claim (&amp;gt;1000 mg/dL) in contrary to haemoglobin (&amp;gt;0.3 g/dL). No interference was detected for bilirubin with concentrations up to 59.8 mg/dL. Conclusion The accuracy of the T50 test was high with nearly identical T50 results between two laboratory sites. Furthermore, no interference was observed within usual clinical ranges of haemoglobin, bilirubin, and lipid levels. Interference was present at extremely high values only.

#1017 External validation of a machine learning model for the prediction of bioimpedance-determined pre-dialysis overhydration

Abstract Background and Aims An accurate, automated determination of volume status in dialysis patients using a reproducible standard would be clinically useful. We have previously developed a machine learning model for the prediction of pre-dialysis overhydration (OH) index (litres) from the Fresenius Body Composition Monitor (BCM). External validation is a requisite step before the clinical implementation of a machine learning model. Method Data analyses were performed using Python. An XGBoost regressor model was internally developed and validated in a Fresenius dataset of 9,452 patients and 279,324 BCM target events from the Czech Republic and Italy between January 2016 and December 2022. An XGBoost regressor model with 60 input variables including the average BCM OH value in the preceding year, 2-week moving averages of dialysis session data and biomarkers/ medications data from the preceding 13 weeks demonstrated the best performance of the models tested. The external validation (EV) dataset consisted of 24 patients from a 7-week observational study in a single centre, in Beaumont, Ireland comprising 94 BCM sessions and 650 haemodialysis sessions. Individualised predictions were compared with BCM measures and clinical volume assessments performed at 2-week intervals. Mean squared error (MSE), Root mean square error (RMSE), and Bland-Altman plots were used for continuous outcomes. Precision, recall and F1 score were used to assess the algorithm's performance in predicting the following fluid categories: overhydration ≥1.1L, normohydration −1.1L to 1.1L, underhydration ≤-1.1L. Normohydration weight was calculated by subtracting the predicted OH index from the pre-dialysis weight. Results There was performance degradation between the internal validation (IV) and EV set. The RMSE for pre-dialysis OH was 1.04 kg in the IV set and 1.5 kg in the EV set. This could be attributed to differences in the distribution of the outcome variable (BCM OH validation dataset=1.97 ± 1.56 kg*, EV dataset=0.89 ± 1.7 kg), and the case-mix. The RMSE for normohydration weight was 1.86 kg in the EV set. Wide limits of agreement were noted on a Bland Altman plot comparing the average of BCM and predicted values plotted against the difference in values (mean bias - 0.54, 95% limits of agreement -3.44 kg to 2.36 kg). The overall accuracy of the model in predicting fluid classes was 60%. A histogram showed a close alignment in the distributions of predicted normohydration weights and BCM values. Conclusion The performance of the algorithms degraded between the internal validation and external validation datasets. However, the pre-dialysis algorithm showed an ability to discriminate fluid categories compared to nursing staff and the overall accuracy was acceptable. Additional training of the algorithms using fluid status classes would be expected to improve the precision of predictions. *Values less than the 1st percentile of – 2.17 L were excluded.