Limits Of Agreement Research Articles

X-ray Coronary Angiogram images and SYNTAX score to develop Machine-Learning algorithms for CHD Diagnosis

Coronary Heart Disease (CHD) is becoming a leading cause of death worldwide. To assess coronary artery narrowing or stenosis, doctors use coronary angiography, which is considered the gold-standard method. Interventional cardiologists rely on angiography to decide on the best course of treatment for CHD, such as revascularization with bypass surgery, coronary stents, or medication. However, angiography has some issues, including operator bias, inter-observer variability, and poor reproducibility. The automated interpretation of coronary angiography is yet to be developed, and these tasks can only be performed by highly specialized physicians. Developing automated angiogram interpretation and coronary artery stenosis estimation using Artificial Intelligence (AI) approaches requires a large dataset of X-ray angiography images that include clinical information. We have collected 231 X-ray images of heart vessels, along with the necessary angiographic variables, including the SYNTAX score, to support the advancement of research on CHD-related machine learning and data mining algorithms. We hope that this dataset will ultimately contribute to advances in clinical diagnosis of CHD.

Transitioning from stress electrocardiogram to cardiopulmonary exercise testing: a paradigm shift toward comprehensive medical evaluation of exercise function.

Cardiopulmonary exercise testing (CPET) has emerged as a powerful diagnostic tool, providing comprehensive physiological insights into the integrated function of cardiovascular, respiratory, and metabolic systems. Exploiting physiological interactions, CPET allows in-depth diagnostic insights. CPET performance entrains several complexities. Interpreting CPET data can be challenging, requiring significant physiological expertise. The advent of artificial intelligence (AI) has introduced a transformative approach to CPET interpretation, enhancing accuracy, efficiency, and clinical decision-making. This review article explores the current state of AI applications in CPET, highlighting AI's potential to replace the traditional stress electrocardiogram (ECG) test as the preferred diagnostic tool in preventive medicine and medical screening. The article discusses the underlying principles of AI, its integration into CPET interpretation, and the associated benefits, including improved diagnostic accuracy, reduced interobserver variability, and expedited decision-making. Additionally, it addresses the challenges and considerations surrounding the implementation of AI in CPET such as data quality, model interpretability, and ethical concerns. The review concludes by emphasizing the significant promise of AI-assisted CPET interpretation in revolutionizing preventive medicine and medical screening settings and enhancing patient care.

Analysing diagnostic practices and referral pathways for glaucoma in Australian primary eye care.

Glaucoma is the leading cause of irreversible blindness globally, posing a significant public health challenge in Australia, particularly among individuals aged 55 years and older. As primary health care providers, optometrists play a crucial role in the early diagnosis and management of glaucoma, making them central to efforts aimed at reducing the burden of this sight-threatening condition. This study investigates the practice patterns of Australian optometrists in diagnosing and managing glaucoma, focusing on test utilisation, diagnostic confidence, referral practices and intra- and inter-observer variabilities in grading glaucomatous optic neuropathy (GON). A mixed-method cross-sectional design was conducted, involving 50 Australian optometrists who graded 120 colour digital retinal photographs for GON and completed an online survey regarding their diagnostic methods and confidence levels. Statistical analyses assessed inter- and intra-observer agreement in GON grading. The results showed that 82% of optometrists surveyed possessed optical coherence tomography (OCT) instruments and 96% had visual field analysers. Despite a majority expressing confidence in glaucoma detection, only 8% felt capable of independently diagnosing the disease and initiating treatment. Inter-observer agreement for glaucoma detection from retinal photographs was moderate (kappa = 0.53, 95% CI 0.50-0.54), while intra-observer agreement was substantial (kappa = 0.73, 95% CI 0.70-0.77). Inter-observer agreement of optometrists was similar to that of ophthalmologists. Most optometrists have access to advanced diagnostic tools, know how to appropriately diagnose and manage glaucoma and have similar inter-observer variability when assessing fundus photographs to that of glaucoma sub-specialists, but few feel confident in independently diagnosing and managing glaucoma.

Semantic segmentation for individual thigh skeletal muscles of athletes on magnetic resonance images.

The skeletal muscles that athletes should train vary depending on their discipline and position. Therefore, individual skeletal muscle cross-sectional area assessment is important in the development of training strategies. To measure the cross-sectional area of skeletal muscle, manual segmentation of each muscle is performed using magnetic resonance (MR) imaging. This task is time-consuming and requires significant effort. Additionally, interobserver variability can sometimes be problematic. The purpose of this study was to develop an automated computerized method for semantic segmentation of individual thigh skeletal muscles from MR images of athletes. Our database consisted of 697 images from the thighs of 697 elite athletes. The images were randomly divided into a training dataset (70%), a validation dataset (10%), and a test dataset (20%). A label image was generated for each image by manually annotating 15 object classes: 12 different skeletal muscles, fat, bones, and vessels and nerves. Using the validation dataset, DeepLab v3+ was chosen from three different semantic segmentation models as a base model for segmenting individual thigh skeletal muscles. The feature extractor in DeepLab v3+ was also optimized to ResNet50. The mean Jaccard index and Dice index for the proposed method were 0.853 and 0.916, respectively, which were significantly higher than those from conventional DeepLab v3+ (Jaccard index: 0.810, p < .001; Dice index: 0.887, p < .001). The proposed method achieved a mean area error for 15 objective classes of 3.12%, useful in the assessment of skeletal muscle cross-sectional area from MR images.

Analysis of the factors influencing the proximity and agreement between critical power and maximal lactate steady state: a systematic review and meta-analyses

Identifying the boundary between heavy and severe exercise domains is crucial since it demarcates the transition from sustainable to unsustainable exercise. This systematic review aimed to determine differences and agreement between two indices used to determine this boundary, namely critical power (CP) and maximal lactate steady state (MLSS), and how moderators may affect these differences. Ten out of 782 studies found were included in the meta analyses. Random effect meta-analyses were performed to evaluate the mean differences (MD) between CP and MLSS, and moderators’ effect on MD was assessed using meta-regression. CP and MLSS agreement was tested using Bland-Altman meta-analyses on the limits of agreements (LoA) of the MD. Power output (PO) at CP was higher (MD (95% LoA) = 12.42 [−19.23; 44.08] W, p = 0.005) than PO at MLSS, with no differences between CP and MLSS in terms of oxygen uptake (MD (95% LoA) = 0.09 [−0.34; 0.52] L⋅min−1, p = 0.097), heart rate (MD (95% LoA) = 0.61 [−15.84; 17.05] bpm, p = 0.784), and blood lactate concentration (MD (95% LoA) = 1.63 [−2.85; 6.11] mM, p = 0.240). Intensities at CP (p = 0.002) and MLSS (p = 0.010) influenced the MD expressed in W. In conclusion, solely when expressed in PO, CP was higher than MLSS, with larger differences in fitter and younger individuals, emphasizing the possible effect of the indicators used for assessing exercise intensity. Finally, the high interindividual variability observed in the differences between CP and MLSS could compromise their interchangeability in predicting the heavy to severe boundary regardless of the parameter used to assess exercise intensity.

Reliability of an all-in-one wearable sensor for continuous vital signs monitoring in high-risk patients: the NIGHTINGALE clinical validation study.

Continuous vital signs monitoring with wearable systems may improve early recognition of patient deterioration on hospital wards. The objective of this study was to determine whether the wearable Checkpoint Cardio's CPC12S, can accurately measure heart rate (HR), respiratory rate (RR), oxygen saturation (SpO2), blood pressure (BP) and temperature continuously. In an observational multicenter method comparison study of 70 high-risk surgical patients admitted to high-dependency wards; HR, RR, SpO2, BP and temperature were simultaneously measured with the CPC12S system and with ICU-grade monitoring systems in four European hospitals. Outcome measures were bias and 95% limits of agreement (LoA). Clinical accuracy was assessed with Clarke Error Grid analyses for HR and RR. A total of 3,212h of vital signs data (on average 26h per patient) were analyzed. For HR, bias (95% LoA) of the pooled analysis was 0.0 (-3.5 to 3.4), for RR 1.5 (-3.7 to 7.5) and for SpO2 0.4 (-3.1 to 4.0). The CPC12S system overestimated BP, with a bias of 8.9 and wide LoA (-23.3 to 41.2). Temperature was underestimated with a bias of -0.6 and LoA of -1.7 to 0.6. Clarke Error Grid analyses showed that adequate treatment decisions regarding changes in HR and RR would have been made in 99.2% and 92.0% of cases respectively. The CPC12S system showed high accuracy for measurements of HR. The accuracy of RR, SpO2 were slightly overestimated and core temperature underestimated, with LoA outside the predefined clinical acceptable range. The accuracy of BP was unacceptably low.

Validity of the Fitbit wearable activity monitor to estimate step counts in free-living conditions in ambulatory children and youth living with disability

Purpose To assess the validity of the Fitbit ChargeHR versus a research-grade accelerometer (ActiGraph GT3X) for estimating daily step counts in free-living conditions in ambulatory children and youth living with physical and non-physical disabilities. Materials and Methods Children and youth living with disability (n = 29; median age 10 years (IQR: 8–13), 55% boys; n = 2 with mobility aid) wore the GT3X ActiGraph accelerometer (hip) and the Fitbit ChargeHR (wrist) for seven days. Inter-device agreement in steps/day was assessed by intraclass correlation coefficients (ICCs) and Bland-Altman plots. A receiver operating curve (ROC) was used to determine a Fitbit step-count cut-point that corresponds to meeting physical activity guidelines (defined as ≥60 min of moderate-to-vigorous physical activity per day). Results Overall, we found an ICC = 0.861 (p < 0.001) between daily step counts measured by the two devices. Bland-Altman analyses revealed a mean difference (“bias”) between the devices with the Fitbit recording, on average, 1,388 more steps/day than the accelerometer (Limits of Agreement (LoA) 1,741 to −4,518 steps per day). The ROC revealed a Fitbit cut-point of 12,272 steps/day corresponding to meeting guidelines. Conclusions Fitbit ChargeHR devices tend to overestimate daily step counts, but may still provide useful estimates of step counts and patterns in children and youth living with disability.

Testing an inverse modeling approach with gradient boosting regression for stroke volume estimation using patient thermodilution data

Stroke volume (SV) is a major indicator of cardiovascular function, providing essential information about heart performance and blood flow adequacy. Accurate SV measurement is particularly important for assessing patients with heart failure, managing patients undergoing major surgeries, and delivering optimal care in critical settings. Traditional methods for estimating SV, such as thermodilution, are invasive and unsuitable for routine diagnostics. Non-invasive techniques, although safer and more accessible, often lack the precision and user-friendliness needed for continuous bedside monitoring. We developed a modified method for SV estimation that combines a validated 1-D model of the systemic circulation with machine learning. Our approach replaces the traditional optimization process developed in our previous work, with a regression method, utilizing an in silico-generated dataset of various hemodynamic profiles to create a gradient boosting regression-enabled SV estimator. This dataset accurately mimics the dynamic characteristics of the 1-D model, allowing for precise SV predictions without resource-intensive parameter adjustments. We evaluated our method against SV values derived from the gold standard thermodilution method in 24 patients. The results demonstrated that our approach provides a satisfactory agreement between the predicted and reference data, with a MAE of 16 mL, a normalized RMSE of 21%, a bias of −9.2 mL, and limits of agreement (LoA) of [−47, 28] mL. A correlation coefficient of r = 0.7 (p &amp;lt; 0.05) was reported, with the predicted SV slightly underestimated (68 ± 23 mL) in comparison to the reference SV (77 ± 26 mL). The significant reduction in computational time of our method for SV assessment should make it suitable for real-time clinical applications.

Test-retest reliability and factor structure of the Danish 6-item version of the Hopkins Symptom Checklist-core depression (SCL-6) in adolescents

BackgroundShort efficient questionnaires to detect depressive symptoms in adolescents are sparsely evaluated. We aimed to examine the test–retest reliability and structural and convergent validity of the Danish version of the 6-item version of the Hopkins Symptom Checklist-core depression (SCL-6) in adolescents.MethodsOur study population consisted of 122 adolescents. Ninety-one adolescents completed SCL-6 (test sample) in the first round, 82 adolescents completed SCL-6 in the second round (retest sample), with 66 completing the questionnaire both rounds (test–retest sample). Reliability was evaluated by intraclass correlation (ICC) and standard error of measurement (SEM), and structural validity was evaluated by confirmatory factor analysis (CFA) and Mokken analyses. Convergent validity was assessed using the 4-item version of the Centre for Epidemiological Studies Depression Scale (CES-DC4).ResultsThe mean sum score in the test sample was higher than in the retest sample (0.74 (95% CI:0.06;1.43)). The limits of agreement (LoA) were broad (-4.73;6.21). The intraclass correlation (ICC(1,1)) was 0.79 (95% CI:0.67;0.87)) showed good reliability, while the SEM (2.03 (95% CI: 1.68; 2.37)) was large considering the range of the scale. Cronbach's alpha showed good internal consistency at both test (0.80) and retest (0.81). The inter-item correlations and Mokken analyses supported the conclusion of a unidimensional scale. The CFA did not find an acceptable fit of a one factor solution. The convergent validity showed a moderate correlation with the CES-DC4 (0.48).ConclusionSCL-6 showed acceptable internal consistency and test–retest reliability. The results from the CFA were inconclusive in terms of demonstrating unidimensionality, but Mokken analyses supported unidimensionality like previous studies. We find the scale usable for population-based research on depressive symptoms in adolescents, but do not recommend it as screening tool on individual level.

The Relevance of the Technical Error of Measurement in Anthropometry: A Pilot Study from The Nutrifunction Project

The Relevance of the Technical Error of Measurement in Anthropometry: A Pilot Study from The Nutrifunction Project

Comparison of noninvasive electrical cardiometry and transpulmonary thermodilution for cardiac output measurement in critically ill patients: a prospective observational study

BackgroundCardiac output (CO) monitoring is essential for diagnosing and managing critically ill patients. Recently, a non-invasive haemodynamic monitoring technique, electrical cardiometry (EC), has gathered increasing interest among ICU physicians. This study aimed to explore the accuracy of CO estimated by non-invasive EC (COEC) compared to CO determined by transpulmonary thermodilution (COTPTD) and to evaluate the ability of COEC to track COTPTD changes (ΔCOTPTD).MethodsThis prospective, observational, single-center study was conducted from April 2021 to April 2023, involving patients who required haemodynamic monitoring using a transpulmonary thermodilution device (PiCCO). COTPTD and COEC were recorded simultaneously, with the investigators obtaining the COEC measurements were blinded to the COTPTD results and vice versa. Agreement between the methods was evaluated using Bland–Altman analysis and percentage error (PE). The ability of COEC to track changes in COTPTD was examined using four-quadrant and polar plots.ResultsSeventy-two patients with PiCCO haemodynamic monitoring were included, yielding 285 paired CO measurements. The bias between COEC and COTPTD was 0.47 L/min, with a limit of agreement (LoA) ranging from -2.91 to 3.85 L/min and a PE of 54.0%. Among 212 pairs of ΔCO data, excluding a central zone of 15% in the four-quadrant plot, the concordance rate between ΔCOEC % and ΔCOTPTD % was 70%. In the polar plot, excluding a central zone with a radius of 0.625 L/min (10% of the mean COTPTD), the mean polar angle for ΔCOEC was 2.2°, with a radial LoA of 56.0°. Exploratory subgroup analysis indicated a PE of 47.0% between COEC and COTPTD and a concordance rate of 72% between ΔCOEC% and ΔCOTPTD% in patients with normal CO (CO ≥ 4 L/min). In patients with elevated thoracic fluid content (TFC > 35 kΩ), the PE between COEC and COTPTD was 45.0%, with a concordance rate of 64% between ΔCOEC% and ΔCOTPTD%. Additionally, in patients receiving low-dose norepinephrine equivalents (NEE ≤ 0.25 μg/kg/min), COEC and COTPTD exhibited a PE of 45.0%, while ΔCOEC% and ΔCOTPTD% achieved a concordance rate of 75% and a radial LoA of 44.2°.ConclusionIn critically ill patients, non-invasive EC indicated limited accuracy in measuring CO, along with a restricted ability to reliably track CO changes. These findings suggested that EC may not be interchangeable with TPTD in the general ICU population.

Interobserver Variability Regarding Adequacy of Oropharyngeal Exposure for Transoral Robotic Surgery May Necessitate Staging Endoscopy: Evaluating TORS Exposure.

Interobserver Variability Regarding Adequacy of Oropharyngeal Exposure for Transoral Robotic Surgery May Necessitate Staging Endoscopy: Evaluating TORS Exposure.

PET/CT and MR Improve Interobserver Agreement in Primary Tumor Determination for Radiotherapy in Esophageal Squamous Cell Cancer

Background/Objectives: The aim of the study was to evaluate interobserver variability in the determination of the primary tumor for radiotherapy treatment planning in esophageal squamous cell carcinoma (ESCC). Methods: Sixteen patients with locally advanced ESCC were included in the analysis. In all patients positron emission tomography with computed tomography (PETC/CT) and magnetic resonance (MR) scans for radiotherapy planning were performed. Five experienced radiation oncologists delineated the primary tumor based on CT alone, MR alone, PET/CT, CT with fused MR and PET/CT with fused MR. Mean tumor volumes were calculated for each patient and imaging modality. The generalized conformity index (CIgen) was calculated to assess agreement in tumor determination. Results: The mean tumor volumes and CIgen for CT alone, MR alone, PET/CT, CT with fused MR and PET/CT with fused MR were 33.1 cm3, 30.2 cm3, 38.1 cm3, 31.9 cm3, 36.2 cm3 and 0.59, 0.64, 0.66, 0.63, 0.71, respectively. CIgen was significantly higher using PET/CT with fused MR compared to CT (p &lt; 0.001) and PET/CT (p = 0.002) and using PET/CT compared to CT (alone) (p = 0.003). Conclusions: Our study showed higher agreement in primary tumor determination in ESCC using PET/CT compared to CT alone. Higher agreement was also found using PET/CT with fused MR compared to CT alone and PET/CT.

Reliability of Self-monitoring of Intraocular Pressure with iCare Home2 Rebound Tonometry.

Using iCare Home2 (iCare, Finland) rebound tonometry, self-measurement of intraocular pressure has demonstrated good reliability and ease of use. To investigate reliability and repeatability of self-measured intraocular pressure (IOP) with rebound tonometry using iCare Home2. 104 patients out of 110 consecutive patients were recruited for this observational cross-sectional study. One randomly selected eye from each patient underwent six consecutive IOP measurements with Goldmann applanation tonometry (GAT), iCare IC200 and iCare Home2. Every eye was tested twice with each device, in random order, by an ophthalmologist for GAT and IC200, and by the patient itself for Home2. In addition, central corneal thickness (CCT) has been collected. Reliability of Home2 has been tested calculating Limits of Agreement (LoA) between self-measured and physician-measured IOP, using the Bland-Altman analysis. Repeatability of each device has been tested calculating Limits of Repeatability (LoR) with the same method. Pearson correlation coefficient was used to determine the correlation between differences in IOP measurements and CCT. The mean difference between GAT and iCare Home2 was -0.28±1.57mm Hg (P=0.070), 95%-LoA: [-3.36, 2,79mm Hg]. The mean difference between IC200 and iCare Home2 was 0.92±1.48mm Hg (P<0.0001), 95%-LoA [-1.98, 3.82mm Hg]. The mean difference between the first and second measurement with GAT, iCare IC200, and iCare Home2 measurements was 0.21±0.98mm Hg (P=0.03), -0.02±1.11mm Hg (P=0.88) and -0.23±1.04mm Hg (P=0.05). Self-measured IOP with rebound tonometry showed good reliability and repeatability when compared with physician-measured IOP with both standard GAT and rebound tonometry.

Calcaneal Ultrasound Attenuation: Does the Region of Interest and Loading Influence the Repeatability of Measurement?

Current calcaneal quantitative ultrasound systems assess different regions of interest (ROI), under different levels of lower limb loading, yield different parameter values, and are likely prone to different levels of error. This study evaluated the repeatability of measures of frequency-dependent attenuation (FDA, 0.3–0.8 MHz) at three calcaneal ROI, Brooke–Wavell (BW), Jaworski (JA), and foot gauge (FG), under four loading conditions (non-weightbearing, semi-weightbearing, bipedal stance, and unipedal stance). FDA in the calcaneus was assessed in 20 healthy participants (mean (± SD) age, 41.7 ± 19.6 years; height, 1.70 ± 0.16 m; and weight, 70.1 ± 23.0 kg) using a custom-built transmission-mode ultrasound system. Reliability was evaluated using the standard error of measurement (SEM) and limits of agreement (LA) and tolerance (95%TL). Differences in mean FDA values between ROI, loading, and measurement occasions were assessed using a repeated measures ANOVA (α = .05). Mean FDA values ranged between 58.0 ± 32.0 and 77.2 ± 27.6 dB/MHz across all conditions. Repeatability of FDA was dependent on the ROI examined and tended to improve with weightbearing. The narrowest limits for 95%TL ranged between ± 15.1 dB/MHz (JA SWB) and ± 62.7 dB/MHz (BW NWB) across sites. The SEM was approximately 10 dB/MHz for both FG and JA during non-weightbearing and was reduced to around 5 dB/MHz with full weightbearing. This study demonstrates that, although measures of ultrasound FDA are dependent on the ROI, lower limb loading may be a useful method to improve the repeatability of FDA measurements.

Evaluating Creativity: Can LLMs Be Good Evaluators in Creative Writing Tasks?

The evaluation of creative writing has long been a complex and subjective process, made even more intriguing by the rise of advanced Artificial Intelligence (AI) tools like Large Language Models (LLMs). This study evaluates the potential of LLMs as reliable and consistent evaluators of creative texts, directly comparing their performance with traditional human evaluations. The analysis focuses on key creative criteria, including fluency, flexibility, elaboration, originality, usefulness, and specific creativity strategies. Results demonstrate that LLMs provide consistent and objective evaluations, achieving higher Inter-Annotator Agreement (IAA) compared with human evaluators. However, LLMs face limitations in recognizing nuanced, culturally specific, and context-dependent aspects of creativity. Conversely, human evaluators, despite lower consistency and higher subjectivity, exhibit strengths in capturing deeper contextual insights. These findings highlight the need for the further refinement of LLMs to address the complexities of creative writing evaluation.

Real-World Application of Digital Morphology Analyzers: Practical Issues and Challenges in Clinical Laboratories

Digital morphology (DM) analyzers have advanced clinical hematology laboratories by enhancing the efficiency and precision of peripheral blood (PB) smear analysis. This review explores the real-world application of DM analyzers with their benefits and challenges by focusing on PB smear analysis and less common analyses, such as bone marrow (BM) aspirates and body fluids (BFs). DM analyzers may automate blood cell classification and assessment, reduce manual effort, and provide consistent results. However, recognizing rare and dysplastic cells remains challenging due to variable algorithmic performances, which affect diagnostic reliability. The quality of blood film as well as staining techniques significantly influence the accuracy of DM analyzers, and poor-quality samples may lead to errors. In spite of reduced inter-observer variability compared with manual counting, an expert’s review is still needed for complex cases with atypical cells. DM analyzers are less effective in BM aspirates and BF examinations because of their higher complexity and inconsistent sample preparation compared with PB smears. This technology relies heavily on artificial intelligence (AI)-based pre-classifications, which require extensive, well-annotated datasets for improved accuracy. The performance variation across platforms in BM aspirates and rare-cell analysis highlights the need for AI algorithm advancements and DM analysis standardization. Future clinical practice integration will likely combine advanced digital platforms with skilled oversight to enhance diagnostic workflow in hematology laboratories. Ongoing research aims to develop robust and validated AI models for broader clinical applications and to overcome the current limitations of DM analyzers. As technology evolves, DM analyzers are set to transform laboratory efficiency and diagnostic precision in hematology.

Intraplaque haemorrhage quantification and molecular characterisation using attention based multiple instance learning.

Intraplaque haemorrhage (IPH) represents a critical feature of plaque vulnerability as it is robustly associated with adverse cardiovascular events, including stroke and myocardial infarction. How IPH drives plaque instability is unknown. However, its identification and quantification in atherosclerotic plaques is currently performed manually, with high inter-observer variability, limiting its accurate assessment in large cohorts. Leveraging the Athero-Express biobank, an ongoing study comprising a comprehensive dataset of histological, transcriptional, and clinical information from 2,595 carotid endarterectomy patients, we developed an attention-based additive multiple instance learning (MIL) framework to automate the detection and quantification of IPH across whole-slide images of nine distinct histological stains. We demonstrate that routinely available Haematoxylin and Eosin (H&E) staining outperformed all other plaque relevant Immunohistochemistry (IHC) stains tested (AUROC = 0.86), underscoring its utility in quantifying IPH. When combining stains through ensemble models, we see that H&E + CD68 (a macrophage marker) as well as H&E + Verhoeff-Van Gieson elastic fibers staining (EVG) leads to a substantial improvement (AUROC = 0.92). Using our model, we could derive IPH area from the MIL-derived patch-level attention scores, enabling not only classification but precise localisation and quantification of IPH area in each plaque, facilitating downstream analyses of its association and cellular composition with clinical outcomes. By doing so, we demonstrate that IPH presence and area are the most significant predictors of both preoperative symptom presentation and major adverse cardiovascular events (MACE), outperforming manual scoring methods. Automating IPH detection also allowed us to characterise IPH on a molecular level at scale. Pairing IPH measurements with single-cell transcriptomic analyses revealed key molecular pathways involved in IPH, including TNF-α signalling, extracellular matrix remodelling and the presence of foam cells. This study represents the largest effort in the cardiovascular field to integrate digital pathology, machine learning, and molecular data to predict and characterize IPH which leads to better understanding how it drives symptoms and MACE. Our model provides a scalable, interpretable, and reproducible method for plaque phenotyping, enabling the derivation of plaque phenotypes for predictive modelling of MACE outcomes.

A Flexible PVDF Sensor for Forcecardiography.

Forcecardiography (FCG) uses force sensors to record the mechanical vibrations induced on the chest wall by cardiac and respiratory activities. FCG is usually performed via piezoelectric lead-zirconate titanate (PZT) sensors, which simultaneously record the very slow respiratory movements of the chest, the slow infrasonic vibrations due to emptying and filling of heart chambers, the faster infrasonic vibrations due to movements of heart valves, which are usually recorded via Seismocardiography (SCG), and the audible vibrations corresponding to heart sounds, commonly recorded via Phonocardiography (PCG). However, PZT sensors are not flexible and do not adapt very well to the deformations of soft tissues on the chest. This study presents a flexible FCG sensor based on a piezoelectric polyvinylidene fluoride (PVDF) transducer. The PVDF FCG sensor was compared with a well-assessed PZT FCG sensor, as well as with an electro-resistive respiratory band (ERB), an accelerometric SCG sensor, and an electronic stethoscope for PCG. Simultaneous recordings were acquired with these sensors and an electrocardiography (ECG) monitor from a cohort of 35 healthy subjects (16 males and 19 females). The PVDF sensor signals were compared in terms of morphology with those acquired simultaneously via the PZT sensor, the SCG sensor and the electronic stethoscope. Moreover, the estimation accuracies of PVDF and PZT sensors for inter-beat intervals (IBIs) and inter-breath intervals (IBrIs) were assessed against reference ECG and ERB measurements. The results of statistical analyses confirmed that the PVDF sensor provides FCG signals with very high similarity to those acquired via PZT sensors (median cross-correlation index of 0.96 across all subjects) as well as with SCG and PCG signals (median cross-correlation indices of 0.85 and 0.80, respectively). Moreover, the PVDF sensor provides very accurate estimates of IBIs, with R2 > 0.99 and Bland-Altman limits of agreement (LoA) of [-5.30; 5.00] ms, and of IBrIs, with R2 > 0.96 and LoA of [-0.510; 0.513] s. The flexibility of the PVDF sensor makes it more comfortable and ideal for wearable applications. Unlike PZT, PVDF is lead-free, which increases safety and biocompatibility for prolonged skin contact.

Systematic review and meta-analysis of artificial intelligence in classifying HER2 status in breast cancer immunohistochemistry.

The DESTINY-Breast04 trial has recently demonstrated survival benefits of trastuzumab-deruxtecan (T-DXd) in metastatic breast cancer patients with low Human Epidermal Growth Factor Receptor 2 (HER2) expression. Accurate differentiation of HER2 scores has now become crucial. However, visual immunohistochemistry (IHC) scoring is labour-intensive and prone to high interobserver variability, and artificial intelligence (AI) has emerged as a promising tool in diagnostic medicine. We conducted a diagnostic meta-analysis to evaluate AI's performance in classifying HER2 IHC scores, demonstrating high accuracy in predicting T-DXd eligibility, with a pooled sensitivity of 0.97 [95% CI 0.96-0.98] and specificity of 0.82 [95% CI 0.73-0.88]. Meta-regression revealed better performance with deep learning and patch-based analysis, while performance declined in externally validated and those utilising commercially available algorithms. Our findings indicate that AI holds promising potential in accurately identifying HER2-low patients and excels in distinguishing 2+ and 3+ scores.