Area Under The Receiver Operating Characteristic Research Articles

Delineation of Groundwater Potential Using the Bivariate Statistical Models and Hybridized Multi-Criteria Decision-Making Models

Identifying groundwater potential zones in a basin and developing a sustainable management plan is becoming more important, especially where surface water is scarce. The main aim of the study is to prepare the groundwater potential maps (GWPMs) considering the bivariate statistical models of frequency ratio (FR), weight of evidence (WoE), and the multi-criteria decision-making (MCDM) model of Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) hybridized with FR and WoE. Two distance measures, Euclidean and Manhattan, were used in TOPSIS to evaluate their effect on GWPMs. The research focused on the Burdur Lake catchment located in the southwest of Türkiye. In total, 74 wells with high yields were chosen randomly for the analysis, 52 (70%) for training, and 22 (30%) for testing processes. Sixteen groundwater conditioning factors were selected. The area under the receiver operating characteristic (AUROC) and true skill statistics (TSS) were utilized to examine the goodness-of-fit and prediction accuracy of approaches. The TOPSIS-WoE-Manhattan model and the FR and WoE models gave the best AUROC values of 0.915 and 0.944 for the training and testing processes, respectively. The best TSS values of 0.827 and 0.864 were obtained by the TOPSIS-FR-Euclidean and WoE models for the training and testing processes, respectively.

Abstract B035: Detection and monitoring of translocation renal cell carcinoma via epigenomic profiling of cell-free DNA

Abstract We evaluated the ability of plasma chromatin immunoprecipitation and sequencing (ChIP-seq) and methylated DNA immunoprecipitation sequencing (MeDIP-seq) to detect translocation renal cell carcinoma (tRCC), to distinguish tRCC from clear cell renal cell carcinoma (ccRCC), and to monitor the disease evolution. tRCC is an aggressive subtype of kidney cancer usually driven by a fusion involving the TFE3 gene. Due to histologic overlap with other subtypes of RCC, tRCC is frequently misclassified. Methods for accurate diagnosis and detection of this molecularly distinct entity are therefore a pressing need. Most cell-free DNA (cfDNA) technologies are genomic-based and not optimized for tRCC as molecular fusions are not easily detected in circulating tumor DNA (ctDNA), and 50% of tRCC cases do not have other alteration. Epigenomic profiling of ctDNA via ChIP-seq has recently emerged as a powerful tool to detect and molecularly subtype cancers and may offer a more sensitive and specific detection of fusion- driven tumors. We first identified differentially methylated regions (DMRs) and regulatory elements (REs) specific to tRCC vs. ccRCC via MeDIP-seq and H3K4me3/H3K27Ac ChIP-Seq, of 4 tRCC and 5 ccRCC cell lines. We also identified TFE3 transcription factor binding sites (TFBS) via TFE3 ChIP-seq in 3 tRCC cell lines. We collected 30 plasma samples from metastatic patients with tRCC, 12 with ccRCC and 9 healthy individuals. Ultra low pass whole genome sequencing (ulpWGS) was performed to infer ctDNA fraction (TF), and cfMeDIP-seq, H3K4me3/H3K27ac cfChIP-seq for epigenomic profiling. Signal at tRCC-specific regions derived from cell lines profiling was aggregated for each mark and normalized to common active REs/methylated regions and used to discriminate tRCC vs. ccRCC vs. healthy plasma. Classification performance was evaluated using the area under the receiver operating characteristic (AUROC) curve. Overall 10/30 tRCC and 5/12 ccRCC samples had >3% TF by ulpWGS. H3K4me3 cfChIP-seq signal was significantly higher in all tRCC samples compared to healthy plasma (AUROC=1) at tRCC-specific REs and TFE3 TFBS. Furthermore, H3K27ac and H3K4me3 cfChIP-seq signal in plasma were also significantly higher at tRCC-specific REs and TFE3 TFBS in tRCC samples compared to ccRCC (AUROC=0.86-0.89). A cfDNA-based classifier combining the four scores distinguished tRCC from ccRCC with an AUROC=0.94. MeDIP-seq could not discriminate between tRCC and ccRCC. Moreover, three patients had multiple time points, and the integrated cfChIP H3K4me3/H3K27Ac score was able to accurately monitor the evolution of the disease. In conclusion, although a majority of tRCC plasma samples profiled had TF <3%, all could be distinguished from healthy samples on the basis of cfChIP. The combined score of H3K27ac and H3K4me3 cfChIP-Seq was also discriminatory for tRCC vs. ccRCC. Epigenomic profiling of cfDNA appears as a powerful tool for both detection of tRCC and discrimination from ccRCC/healthy plasma, with potential implications for diagnosis, disease monitoring and guiding therapy. Citation Format: Simon Garinet, Karl Semaan, John Caniff, Noa Phillips, Jacob Berchuck, Jiao Li, Kevin Lyons, Ananthan Sadagopan, Brad Fortunato, Mary Lee, Jack Horst, Rachel Trowbridge, Ji-Heui Seo, Matthew Freedman, Toni Choueiri, Sylvan Baca, Srinivas Viswanathan. Detection and monitoring of translocation renal cell carcinoma via epigenomic profiling of cell-free DNA [abstract]. In: Proceedings of the AACR Special Conference: Liquid Biopsy: From Discovery to Clinical Implementation; 2024 Nov 13-16; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2024;30(21_Suppl):Abstract nr B035.

Abstract Su803: Sudden Cardiac Arrest Prediction via Deep Learning Electrocardiogram Analysis

Background: Sudden cardiac arrest (SCA) is a potentially fatal event that often occurs without prior indications. Hypothesis: It was hypothesized that electrocardiogram (ECG) data could be used in conjunction with Deep Learning (DL) algorithms to predict the occurrence of SCA’s. Aims: To improve SCA outcomes and enable preventative strategies, the ECG data from the Nightingale Open Science - Subtyping Cardiac Arrest dataset were utilized to create a DL model to serve as a screening tool for sudden cardiac arrest. Methods: A publicly-available dataset containing ten seconds of 12-lead ECGs from individuals who had a SCA and persons who did not have a SCA, and information about time from ECG to arrest, and age and sex were utilized for individualized prediction of SCA. Subsets of the initial data were used to train, validate, and test deep convolution neural network models that processed cardiac cycles from single-lead ECG voltage waveforms. Results: The cohort comprised 221 (mean age: 68.5 and #females: 75) patients who experienced SCA and 1046 controls (mean age: 69.7 and #females: 497) who did not have a cardiac arrest. The SCA ECG-DL base model, with ECGs within one day and with age and sex data, had an area under the receiver operating curve (AUROC) of 0.774 during hold-out testing for discrimination of subsequent SCA. With sensitivity set at 95%, the base model specificity of the SCA ECG-DL model was 31%. Gradient-weighted class activation mapping showed that the SCA ECG-DL model relied most heavily on the QRS complex to make predictions. The performance of the SCA ECG-DL model was similar removing age and sex data (AUROC 0.775) and less reliable when predicting SCA within one month (AUROC 0.734) or between one month and one year (AUROC 0.574). Conclusion: DL interpretation of ECG data is a promising means of screening for SCA, and this method explains differences in SCA’s due to age and sex. Model performance improved when ECGs were obtained at shorter intervals to the incident SCA event, and SCA prediction was more dependent upon QRS complex data than other temporal segments of the ECG.

Abstract 4130756: Two Leads are All We Need: Optimizing Choice of Electrocardiogram Leads for Artificial Intelligence-based Based Detection of Left Ventricular Systolic Dysfunction

Introduction: Artificial intelligence (AI) models built with convolutional neural networks (CNN) accurately detect left ventricular systolic dysfunction (LVSD) from 12-lead electrocardiogram (ECG). While AI models can also detect LVSD from 1-lead (but with lower accuracy), the optimal numbers and combinations of the leads remain unclear. Aims: To identify the optimal numbers and combinations of ECG leads for detecting LVSD. Methods: A total of 75,033 ECGs recorded within 14 days of an echocardiogram were collected. The data was randomly divided into derivation, validation, and test sets in a 5:2:3 ratio without patient overlap. The same split was used throughout the study. While all available ECGs were included in derivation and validation sets, the test set included 1 ECG per patient closest to the echocardiogram. CNN models were trained to detect LVSD, defined as left ventricular ejection fraction <40%, using all possible combinations of 1-, 2-, and 3-lead extracted from 12-lead ECG. The final model for each lead combination was chosen according to the area under the receiver operating curve (AUROC) on the validation set across the 20 epochs. The models were trained 4 times to evaluate the variance caused by initialization vectors. Results: The 12-lead model showed AUROC of 0.893 (95%CI, 0.887-0.899). Multiple models trained with 2-lead achieved similar performance, with aVR-V4 displaying the highest AUROC (0.892; 95%CI, 0.889-0.895), followed by I-V4 (0.889; 95%CI, 0.883-0.895) and I-II (0.888; 95%CI, 0.884-0.893). The results were similar for 3-lead (aVR-aVL-V4: AUROC 0.895; 95%CI, 0.893-0.897, I-aVR-V4: 0.892; 95%CI, 0.889-0.896, and I-II-V4: 0.891; 95%CI, 0.886-0.895). The choice of leads greatly affected the performance. For the 1-lead models, aVR showed the highest AUROC (0.876; 95%CI, 0.873-0.879), but III only achieved an AUROC of (0.766; 95%CI, 0.761-0.772). The combinations, including a chest lead and a limb lead (especially aVR), tend to achieve higher performance (Fig 1). Conclusion: A combination of 2-lead was sufficient to achieve comparable accuracy as the 12-lead for the AI model to detect LVSD. Wearable devices collecting a combination of a limb and a chest lead may enhance the detection of LVSD.

Abstract 4125281: Detection of Acute Myocardial Infarction and Ischemia From Lead-I ECG Using Deep Learning

Introduction: While smartwatches that acquire Lead-I electrocardiogram (ECG) have shown promise in detecting arrhythmias, the detection of ischemic events using Lead-I alone remains understudied. We aim to build and evaluate a deep-learning model to determine whether Lead-I ECG can identify ischemia/infarction in a 12-lead ECG. Methods: We developed a deep-learning model that can identify MI/ischemia using only Lead-I ECG signal as input. For this task, we utilized PTB-XL, a large publicly available dataset of 12-Lead clinical ECG with associated diagnostic labels. Our experiment utilized 8781 ECGs whose labels were validated by at-least one cardiologist with 100% confidence scores. This dataset includes 1391 ECGs that are labeled as either ischemic ST/T changes or MI, and 7390 ECGs with either “normal” labels or non-ischemic ST/T abnormalities. We split the dataset in training, validation, and holdout sets (80-10-10) with equal stratification of the diagnosis classes. The holdout test set contains 909 non-ischemic ECGs and 233 ECGs with MI/ischemia diagnosis, 106 of which are labeled as acute MI or ischemic ST/T changes and 127 labeled as other MI chronicity. The discriminatory ability of the model is evaluated with the Area Under the receiver-operating Curve (AUC), sensitivity, specificity, and the predictive values. We further analyze the model’s capability in discriminating the cohort of acute MI and ischemia diagnoses from other MI as well as from the non-ischemic cohort. Results: The model has an AUC of 0.88 for discriminating an MI/ischemia vs normal. Using a threshold corresponding to 80% sensitivity, the corresponding specificity is 79% and the negative and positive predictive values are 94% and 50% (Table-1). Although the model was trained to identify the combined class of MI/ischemia, it can differentiate acute MI and ischemia cohorts from other MI as well as from non-ischemic diagnoses with p-values<10 -4 (Table-2). Lastly, with the same threshold, the sensitivity for detecting acute inferior myocardial infarctions or inferior ischemia – a class that is difficult to detect with Lead-I alone – is 82%. Conclusions: Our deep learning model, which was trained on Lead-I ECG alone, can effectively detect acute MI and ischemia that would have been identified using a 12-lead ECG. This method presents a proof-of-concept for applications in home settings. Further prospective validation using wearable ECG monitors that record lead-I is warranted.

Abstract 4135928: Externally Validated Deep Learning Model for Patent Ductus Arteriosus Detection by Echocardiography in Preterm Infants

Introduction: Patent ductus arteriosus (PDA) is a common form of congenital heart disease in premature infants and a source of significant morbidity. Echocardiography is the mainstay modality for diagnosis and monitoring but is resource-intensive and requires expertise for interpretation. Aims: To develop and externally validate a deep learning model to identify PDA in preterm infants by echocardiography. Methods: We trained, validated, and tested a convolutional neural network to detect PDA in raw echocardiogram clips from infants (gestational age <37 weeks) with and without PDAs and no significant additional congenital heart disease at the Medical University of South Carolina (MUSC). For training and validation, we employed a 12-fold cross-validation procedure. Experienced cardiologists defined the ground truth labels. We assessed classification performance on an internal model naïve dataset using the area under the receiver operator curve (AUROC), sensitivity, specificity, positive predictive value, and negative predictive value. For external multicenter testing, we used studies from Children’s Hospital of Orange County (CHOC) and Boston Children’s Hospital (BCH), both of which utilize a different ultrasound system vendor. Results: Training and validation were completed using 1,145 color/color-compare clips (661 clips with PDA and 484 no PDA) from 66 patients. The internal test dataset consisted of 142 clips from 16 patients. The external CHOC and BCH cohorts consisted of 146 clips from 50 patients and 294 clips from 68 patients, respectively. Model performance was similar for internal and external testing (Table 1). Conclusions: Our deep learning model detected the presence of PDA with acceptable performance. This performance generalized well across multiple institutions and ultrasound vendors. This work shows promise for the future development of an automated tool for PDA diagnosis enabling the democratization of pediatric cardiology resources.

Abstract 4135726: Prospective validation and implementation pilot study of an emergency department heart failure risk stratification tool: STRIDE-HF

Background: We previously found the STRIDE-HF emergency department (ED) risk tool accurately predicted risk of a 30-day serious adverse event (SAE), including 30-day mortality, cardiopulmonary resuscitation, intra-aortic balloon pump insertion, intubation, new dialysis, myocardial infarction, or coronary revascularization. Research question: We sought to prospectively validate STRIDE-HF and describe safety of the deployed model in a clinical pilot. We hypothesized that performance would decrease slightly with prospective evaluation and that real-time risk display would contribute to safe disposition decisions. Goals: 1) To prospectively validate KPCARES-HF across 21 community EDs among ED patients with acute heart failure (AHF) from January 1, 2023 – December 31 st , 2023; and 2) to describe outcomes among patients predicted to be very low risk and discharged home in the clinical pilot. Methods: For prospective validation, we report 30-day mortality rates by risk strata. We assess model performance using area under the receiver operator curve (AUROC) and sensitivity at a key clinical threshold. We describe rates of 30-day SAE among patients predicted to be very low risk and discharged after ED or observation unit care in the clinical pilot. Results: There were 13,274 ED patients in the prospective validation; median age was 76, 50.8% were female, and 44.5% were non-White. We found 11.4%, 24.8%, 31.9%, and 31.9% of patients were very low, low, moderate, and high risk, respectively, while 21.6%, 15.2% and 63.0% were discharged, observed, and admitted, respectively. Among discharged patients, 29.2% and 13.3% were moderate or high risk, respectively, and 30-day mortality rates among these patients were 2.2% and 8.4%, respectively. Among 8,363 admitted patients, 27.7% were very low or low risk. Conclusions: STRIDE-HF maintained high predictive accuracy with prospective validation in this diverse, multi-center cohort. Our findings suggest that use of STRIDE-HF might help better align risk with admission decision and can be safely used to assist providers in identifying patients who may be stable for outpatient care.

NIMG-57. EMBEDDED GRAPH DERIVED FROM PSEUDO-RESTING-STATE FUNCTIONAL MRI CAN PREDICT COGNITIVE IMPAIRMENT IN GLIOMA

Abstract Resting-state brain network analyses are of great interest in studying the neurocognition and measuring the functional connectivity (FC) patterns of glioma patients. However, resting-state functional MRI (rs-fMRI) is not assessed in clinical routine due to the scan time and cost. The current study used pseudo-resting-state functional MRI (pseudo-rs-fMRI) derived from dynamic susceptibility contrast (DSC) perfusion MRI to predict cognitive impairment in glioma. 37 glioma patients were enrolled consecutively in the current study with DSC perfusion MRI acquired and neurocognition assessed, where 24 glioma patients also got rs-fMRI collected. The pseudo-rs-fMRI was created by voxel-wise subtracting the Gamma-variate modeled signal of contrast agent bolus from the original DSC perfusion signal. Following the pre-processing of pseudo-rs-fMRI and full rs-fMRI, the functional connectivity network (FCN) was established for each patient. Graph embedding was implemented to learn and extract features of each node in the binarized FCN, and the decision tree classification algorithm was applied to distinguish cognitive impairment. FCNs of 24 patients with DSC perfusion and rs-fMRI acquired were used to train the model, and FCNs of 13 patients with only DSC perfusion MRI acquired were used to test the model. The AUC (area under the receiver operating characteristic curve) was used to evaluate the model’s performance. The similarity and mean-square-difference between average FCNs extracted from rs-fMRI and pseudo-rs-fMRI were 0.6769 and 0.0263, respectively. Classifier trained using 24 FCNs of pseudo-rs-fMRI has an AUC of 0.750 in identifying cognitively impaired patients in the testing cohort. Combining FCNs of both rs-fMRI and pseudo-rs-fMRI to train the classifier resulted in a higher AUC of 0.833 in the testing cohort. To summary, DSC perfusion MRI-derived pseudo-rs-fMRI can be used to predict cognitive impairment in patients with gliomas. External validation would be required.

Development and validation of a predictive scoring model for complications following endoscopic endonasal skull base surgery.

The endoscopic endonasal approach (EEA) has evolved into an established technique in skull base surgery. The authors previously examined 1002 EEA procedures and reported factors associated with postoperative complications. Here they report the development and validation of a scoring model based on risk factors to better predict complications following EEA. The authors developed an optimized EEA scoring model for predicting postoperative complications as evidenced by the area under the receiver operating characteristic (AUROC) curve using their previously published data in addition to data collected from the subsequent 292 EEA procedures from years 2010-2020. The model was built systematically by evaluating the contributions that different variables had on the overall predictive ability of the model. The aim was to design a model containing as few variables as possible for practicality and to facilitate calculation and use at the bedside. The Clavien-Dindo grading system was used to classify complications into grades I-V based on the level of intervention that was required to manage the complication, with grades III-V considered to be higher-grade (i.e., those requiring reoperation or ICU-level care or death). The authors identified 1294 EEA operations performed between July 2010 and July 2020 that met their inclusion criteria. Higher-grade complications were identified following 135 EEA operations. The variables that were ultimately included in the model were age, BMI, operative time, meningioma, chordoma, expanded intradural approach, and nasoseptal flap use. The final model yielded an acceptable AUROC curve of 0.72 and predicted a stepwise increase in the rate of higher-grade complications as the score increased. A score of 0-2 (low) on the grading system was associated with an average complication rate of 5.1%. A score of 3-5 (medium) was associated with an average complication rate of 12.6%. A score of 6 or above (high) was associated with an average complication rate of 26%. This EEA complications scoring model accurately categorizes patients into low-, medium-, and high-risk groups with readily obtained variables. A high score in this complications model does not suggest that a patient is ineligible for surgery, but rather highlights the importance of thorough case selection, operating with caution, and appropriate preoperative counseling.

Development and Validation of a Predictive Model for Maternal Cardiovascular Morbidity Events in Patients With Hypertensive Disorders of Pregnancy.

Hypertensive disorders of pregnancy (HDP) are a major contributor to maternal morbidity, mortality, and accelerated cardiovascular (CV) disease. Comorbid conditions are likely important predictors of CV risk in pregnant people. Currently, there is no way to predict which people with HDP are at risk of acute CV complications. We developed and validated a predictive model for all CV events and for heart failure, renal failure, and cerebrovascular events specifically after HDP. Models were created using the Premier Healthcare Database. The inclusion criteria for the model dataset were delivery with an HDP with discharge from October 1, 2015 to December 31, 2020. Machine learning methods were used to derive predictive models of CV events occurring during delivery hospitalization (Index Model) or during readmission (Readmission Model) using a training set (60%) to estimate model parameters, a validation set (20%) to tune model hyperparameters and select a final model, and a test set (20%) to evaluate final model performance. The total model cohort consisted of 553,658 deliveries with an HDP. A CV event occurred in 6501 (1.2%) of the delivery hospitalizations. Multilabel neural networks were selected for the Index Model and Readmission Model due to favorable performance compared to alternatives. This approach is designed for prediction of multiple events that share risk factors and may cooccur. The Index Model predicted all CV events with area under the receiver operating curve (AUROC) 0.878 and average precision (AP) 0.239 (cerebrovascular events: AUROC 0.941, heart failure: AUROC 0.898, and renal failure: AUROC 0.885). With a positivity threshold set to achieve ≥90% sensitivity, model specificity was 65.0%, 83.5%, 68.6%, and 65.6% for predicting all CV events, cerebrovascular events, heart failure, and renal failure, respectively. CV events within 1 year of delivery occurred in 3018 (0.6%) individuals. The Readmission Model predicted all CV events with AUROC 0.717 and AP 0.022 (renal failure: AUROC 0.748, heart failure: AUROC 0.734, and cerebrovascular events AUROC 0.698). Feature importance analysis indicated that the presence of chronic renal disease, cardiac disease, pulmonary hypertension, and preeclampsia with severe features had the greatest effect on the prediction of CV events. Among individuals with HDP, our multilabel neural network model predicted CV events at delivery admission with good classification and events within 1 year of delivery with fair classification.

Characterisation of Paediatric Neuroblastic Tumours by Quantitative Structural and Diffusion-Weighted MRI

Purpose: To determine the diagnostic accuracy of quantitative diffusion-weighted (DW) MRI apparent diffusion coefficient (ADC) and tumour volumes to differentiate between malignant (neuroblastoma (NB)) and benign types of neuroblastic tumours (ganglioneuroma (GN) and ganglioneuroblastoma (GNB)) using different region-of-interest (ROI) sizes. Materials and Methods: This single-centre retrospective study included malignant and benign paediatric neuroblastic tumours that had undergone DW MRI at diagnosis. The outcome was diagnostic accuracy of the tumour volume from structural and ADC DW MRI, in comparison to histopathology (reference standard). Results: Data from 40 patients (NB, n = 24; GNB, n = 6; GN, n = 10), 18 (45%) females and 22 (55%) males, with a median age at diagnosis of 21 months (NB), 64 months (GNB), and 133 months (GN), respectively, ranging from 0 to 193 months, were evaluated. The area under the receiver operating characteristic (AUROC) curve for ADC for discriminating between neuroblastic tumours’ histopathology for a small ROI was 0.86 (95% CI: 0.75–0.98), and for a large ROI, 0.83 (95% CI: 0.71–0.96). An ADC cut-off value of 1.06 × 10−3 mm2/s was able to distinguish malignant from benign tumours with 83% (68–98%) sensitivity and 75% (95% CI: 54–98%) specificity. Tumour volume was not indicative of malignant vs. benign tumour diagnosis. Conclusions: In this study, both small and large ROIs used to derive ADC DW MRI metrics demonstrated high accuracy to differentiate malignant from benign neuroblastic tumours, with the ADC AUROC for the averaged multiple small ROIs being slightly greater than that of large ROIs, but with overlapping 95% CIs. This should be taken into consideration for standardisation of ROI-related data analysis by international initiatives.

A generalised computer vision model for improved glaucoma screening using fundus images.

Worldwide, glaucoma is a leading cause of irreversible blindness. Timely detection is paramount yet challenging, particularly in resource-limited settings. A novel, computer vision-based model for glaucoma screening using fundus images could enhance early and accurate disease detection. To develop and validate a generalised deep-learning-based algorithm for screening glaucoma using fundus image. The glaucomatous fundus data were collected from 20 publicly accessible databases worldwide, resulting in 18,468 images from multiple clinical settings, of which 10,900 were classified as healthy and 7568 as glaucoma. All the data were evaluated and downsized to fit the model's input requirements. The potential model was selected from 20 pre-trained models and trained on the whole dataset except Drishti-GS. The best-performing model was further trained to classify healthy and glaucomatous fundus images using Fastai and PyTorch libraries. The model's performance was compared against the actual class using the area under the receiver operating characteristic (AUROC), sensitivity, specificity, accuracy, precision and the F1-score. The high discriminative ability of the best-performing model was evaluated on a dataset comprising 1364 glaucomatous discs and 2047 healthy discs. The model reflected robust performance metrics, with an AUROC of 0.9920 (95% CI: 0.9920-0.9921) for both the glaucoma and healthy classes. The sensitivity, specificity, accuracy, precision, recall and F1-scores were consistently higher than 0.9530 for both classes. The model performed well on an external validation set of the Drishti-GS dataset, with an AUROC of 0.8751 and an accuracy of 0.8713. This study demonstrated the high efficacy of our classification model in distinguishing between glaucomatous and healthy discs. However, the model's accuracy slightly dropped when evaluated on unseen data, indicating potential inconsistencies among the datasets-the model needs to be refined and validated on larger, more diverse datasets to ensure reliability and generalisability. Despite this, our model can be utilised for screening glaucoma at the population level.

A multiple biomolecules-based rapid life detection protocol embedded in a rover scientific subsystem for soil sample analysis

The study of whether life exists, is extinct, or not depends on various sophisticated experimental studies, as many different signatures of life can be used. The experimental procedures that can be performed to identify life can be further restricted by time, resources, and mobility constraints. Therefore, any research analyzing the presence of extraterrestrial life must be precise and unambiguous. This research focuses on the objective of the extraterrestrial life detection domain and seeks to provide an efficient protocol that can produce life detection decisions based on empirical data obtained through chemical analysis under time and resource-constrained conditions. While the majority of existing frameworks in this field are designed to identify biomolecules, our goal is to accomplish the same with minimal operational expense and mission complexity. We argue that the thoughtful integration of multiple biomolecular detections with lesser complexity and a robust framework can improve overall mission performance by satisfying the necessary time and resource constraints. In this study, a rapid multiple biomolecules-based life detection protocol (MBLDP-R) from soil samples is developed from scratch and embedded in an operational scientific rover subsystem targeted for planetary analysis missions. The study uses artificial biomolecule samples and simulated extraterrestrial environments to illustrate the suggested protocol’s end-to-end process. First, we list a few significant biomolecules, including lipids, proteins, carbohydrates, nucleic acids, ammonia, and pigments. Then, a weighted qualitative test scoring is applied to sort out the best test method for the finally selected biomolecules which are used as operational analogue to showcase the protocol’s in-situ analysis and decision-making capabilities. Based on the suitable biomolecules, a scientific exploration subsystem is developed, and the implemented protocol is built to perform onboard sample analysis. Evaluation results show that: (1) the proposed MBLDP-R protocol could effectively predict the classes with an average f1-score of 98.65% (macro) and 90.00% (micro), (2) the area under the Receiver Operating Characteristics (AUC-ROC) curve shows the sample categories to be correctly predicted 92% of the time (97% for Extant, 88% for Extinct, and 92% in the case of NPL), and (3) the protocol is time-efficient with an average completion time of 17.60 min, demonstrating the protocol’s rapid nature in detecting biosignatures in soil samples. The research outcome yields useful additional data for related future studies, particularly in the design of scientific frameworks for mission-specific requirements with limited resources while also serving as a reference point for constraint evaluation methods for similar systems.

A blood-based biomarker panel for non-invasive diagnosis of metabolic dysfunction-associated steatohepatitis.

The current diagnosis of metabolic dysfunction-associated steatotic liver disease (MASLD) and its severe form, metabolic dysfunction-associated steatohepatitis (MASH), is suboptimal. Here, we recruited 700 individuals, including 184 from Hong Kong as a discovery cohort and 516 from San Diego, Wenzhou, and Hong Kong as three validation cohorts. A panel of 3 parameters (C-X-C motif chemokine ligand 10 [CXCL10], cytokeratin 18 fragments M30 [CK-18], and adjusted body mass index [BMI]) was formulated (termed N3-MASH), which discriminated patients with MASLD from healthy controls with an area under the receiver operating characteristic (AUROC) of 0.954. Among patients with MASLD, N3-MASH could identify patients with MASH with an AUROC of 0.823, achieving 90.0% specificity, 62.9% sensitivity, and 88.6% positive predictive value. The diagnostic performance of N3-MASH was confirmed in three validation cohorts with AUROC of 0.802, 0.805, and 0.823, respectively. Additionally, N3-MASH identifies patients with MASH improvement with an AUROC of 0.857. In summary, we developed a robust blood-based panel for the non-invasive diagnosis of MASH, which might help clinicians reduce unnecessary liver biopsies.

Predictive Models for the Implementation of Targeted Reproductive Management in Multiparous Cows on Automatic Milking Systems

Targeted reproductive management (TRM) aims to improve the fertility efficiency of the dairy herd by applying group-level management strategies based on expected reproductive performance. Key to the utility of TRM is the accuracy with which an animal's reproductive performance can be predicted. Automatic milking systems (AMS) allow for the collection of data relating to milk quantity, quality, and robot visit behavior throughout the transition period. In addition to this, auxiliary data sources such as rumination and activity monitors, as well as historical cow-level data are often readily available. The utility of this data for the prediction of fertility has not been previously explored. The objective of this study was first, to assess the accuracy with which the likelihood of expression of oestrus between 22 and 65 d in milk (DIM) and conception to first insemination between 22 and 80 DIM could be predicted using data collected by AMS from 1 to 21 DIM. Our second objective was to assess the change in model performance following the addition of 2 auxiliary data sources. Using data derived solely from the AMS (RBT data set) a binary random forest classification model was constructed for both outcomes of interest. The performance of these models was compared with models constructed using AMS data in conjunction with 2 auxiliary sources (RBT+ data set). Expression of oestrus was classified with an area under the receiver operator curve (AUC-ROC) of 0.6 and 0.65, conception to first insemination with an AUC-ROC of 0.56 and 0.62 for the RBT and RBT+ data sets respectively. No statistically significant improvement in classification accuracy was achieved by the addition of auxiliary data sources. This is the first study to report the utility of data collected by AMS for the prediction of reproductive performance. Though the performance described is comparable with previously reported models, their utility for the implementation of TRM is limited by poor classification accuracy within key sub-groups. Of note within this study is the failure of the addition of auxiliary data sources to increase the accuracy of prediction over models built using AMS data alone. We discuss the advantages and limitations the integration of additional data sources imposes on model training and deployment and suggest alternative methods to improve performance while preserving model parsimony.

A novel digital health approach to improving global pediatric sepsis care in Bangladesh using wearable technology and machine learning.

Sepsis is the leading cause of child death globally with low- and middle-income countries (LMICs) bearing a disproportionate burden of pediatric sepsis deaths. Limited diagnostic and critical care capacity and health worker shortages contribute to delayed recognition of advanced sepsis (severe sepsis, septic shock, and/or multiple organ dysfunction) in LMICs. The aims of this study were to 1) assess the feasibility of a wearable device for physiologic monitoring of septic children in a LMIC setting and 2) develop machine learning models that utilize readily available wearable and clinical data to predict advanced sepsis in children. This was a prospective observational study of children with sepsis admitted to an intensive care unit in Dhaka, Bangladesh. A wireless, wearable device linked to a smartphone was used to collect continuous recordings of physiologic data for the duration of each patient's admission. The correlation between wearable device-collected vital signs (heart rate [HR], respiratory rate [RR], temperature [T]) and manually collected vital signs was assessed using Pearson's correlation coefficients and agreement was assessed using Bland-Altman plots. Clinical and laboratory data were used to calculate twice daily pediatric Sequential Organ Failure Assessment (pSOFA) scores. Ridge regression was used to develop three candidate models for advanced sepsis (pSOFA > 8) using combinations of clinical and wearable device data. In addition, the lead time between the models' detection of advanced sepsis and physicians' documentation was compared. 100 children were enrolled of whom 41% were female with a mean age of 15.4 (SD 29.6) months. In-hospital mortality rate was 24%. Patients were monitored for an average of 2.2 days, with > 99% data capture from the wearable device during this period. Pearson's r was 0.93 and 0.94 for HR and RR, respectively) with r = 0.72 for core T). Mean difference (limits of agreement) was 0.04 (-14.26, 14.34) for HR, 0.29 (-5.91, 6.48) for RR, and -0.0004 (-1.48, 1.47) for core T. Model B, which included two manually measured variables (mean arterial pressure and SpO2:FiO2) and wearable device data had excellent discrimination, with an area under the Receiver-Operating Curve (AUC) of 0.86. Model C, which consisted of only wearable device features, also performed well, with an AUC of 0.78. Model B was able to predict the development of advanced sepsis more than 2.5 hours earlier compared to clinical documentation. A wireless, wearable device was feasible for continuous, remote physiologic monitoring among children with sepsis in a LMIC setting. Additionally, machine-learning models using wearable device data could discriminate cases of advanced sepsis without any laboratory tests and minimal or no clinician inputs. Future research will develop this technology into a smartphone-based system which can serve as both a low-cost telemetry monitor and an early warning clinical alert system, providing the potential for high-quality critical care capacity for pediatric sepsis in resource-limited settings.

Exercise-changed gut mycobiome as a potential contributor to metabolic benefits in diabetes prevention: an integrative multi-omics study

ABSTRACT Background The importance of gut microbes in mediating the benefits of lifestyle intervention is increasingly recognized. However, compared to the bacterial microbiome, the role of intestinal fungi in exercise remains elusive. With our established randomized controlled trial of exercise intervention in Chinese males with prediabetes (n = 39, ClinicalTrials.gov:NCT03240978), we investigated the dynamics of human gut mycobiome and further interrogated their associations with exercise-elicited outcomes using multi-omics approaches. Methods Clinical variations and biological samples were collected before and after training. Fecal fungal composition was analyzed using the internal transcribed spacer 2 (ITS2) sequencing and integrated with paired shotgun metagenomics, untargeted metabolomics, and Olink proteomics. Results Twelve weeks of exercise training profoundly promoted fungal ecological diversity and intrakingdom connection. We further identified exercise-responsive genera with potential metabolic benefits, including Verticillium, Sarocladium, and Ceratocystis. Using multi-omics approaches, we elucidated comprehensive associations between changes in gut mycobiome and exercise-shaped metabolic phenotypes, bacterial microbiome, and circulating metabolomics and proteomics profiles. Furthermore, a machine-learning algorithm built using baseline microbial signatures and clinical characteristics predicted exercise responsiveness in improvements of insulin sensitivity, with an area under the receiver operating characteristic (AUROC) of 0.91 (95% CI: 0.85–0.97) in the discovery cohort and of 0.79 (95% CI: 0.74–0.86) in the independent validation cohort (n = 30). Conclusions Our findings suggest that intense exercise training significantly remodels the human fungal microbiome composition. Changes in gut fungal composition are associated with the metabolic benefits of exercise, indicating gut mycobiome is a possible molecular transducer of exercise. Moreover, baseline gut fungal signatures predict exercise responsiveness for diabetes prevention, highlighting that targeting the gut mycobiome emerges as a prospective strategy in tailoring personalized training for diabetes prevention.

Transfer learning classification of suspicious lesions on breast ultrasound: is there room to avoid biopsies of benign lesions?

BackgroundBreast cancer (BC) is the most common malignancy in women and the second cause of cancer death. In recent years, there has been a strong development in artificial intelligence (AI) applications in medical imaging for several tasks. Our aim was to evaluate the potential of transfer learning with convolutional neural networks (CNNs) in discriminating suspicious breast lesions on ultrasound images.MethodsTransfer learning performances of five different CNNs (Inception V3, Xception, Densenet121, VGG 16, and ResNet50) were evaluated on a public and on an institutional dataset (526 and 392 images, respectively), customizing the top layers for the specific task. Institutional images were contoured by an expert radiologist and processed to feed the CNNs for training and testing. Postimaging biopsies were used as a reference standard for classification. The area under the receiver operating curve (AUROC) was used to assess diagnostic performance.ResultsNetworks performed very well on the public dataset (AUROC 0.938–0.996). The direct generalization to the institutional dataset resulted in lower performances (max AUROC 0.676); however, when tested on BI-RADS 3 and BI-RADS 5 only, results were improved (max AUROC 0.792). Good results were achieved on the institutional dataset (AUROC 0.759–0.818) and, when selecting a threshold of 2% for classification, a sensitivity of 0.983 was obtained for three of five CNNs, with the potential to spare biopsy in 15.3%–18.6% of patients.ConclusionIn conclusion, transfer learning with CNNs may achieve high sensitivity and might be used as a support tool in managing suspicious breast lesions on ultrasound images.Relevance statementTransfer learning is a powerful technique to exploit the performances of well-trained CNNs for image classification. In a clinical scenario, it might be useful for the management of suspicious breast lesions on breast ultrasound, potentially sparing biopsy in a non-negligible number of patients.Key PointsProperly trained CNNs with transfer learning are highly effective in differentiating benign and malignant lesions on breast ultrasound.Setting clinical thresholds increased sensitivity.CNNs might be useful as support tools in managing suspicious lesions on breast ultrasound.Graphical

SMART-2D: a novel scoring system to predict non diagnostic exercise stress echocardiography in renal transplant candidates

Abstract Background Exercise stress echocardiography (ESE) is often used for pre-renal transplant risk stratification. Nearly 20% of ESE may be non-diagnostic, resulting in delays to transplant and increased healthcare cost. A simple, reliable tool for predicting non-diagnostic ESE may streamline assessment. Methods Retrospective analysis of 898 kidney transplant candidates between 2013-2020 who underwent exercise stress echocardiography for pre-transplant cardiovascular assessment was performed. The cohort was separated into a derivation and validation set. Multivariable logistic regression identified predictors of non-diagnostic ESE. Covariates associated with non-diagnostic ESE (p<0.10) on univariate analysis were included in the multivariable model. From this model, only variables with p<0.05 were included for score development. Model discrimination was assessed by area under the receiver operating curve (AUROC). Results Non-diagnostic ESE occurred in 17% (151/898). Predictors of non-diagnostic study formed the scoring system: 1 point for female sex, 2 points for BMI >40, 1 point for age >55, Heart Rate<55 4 points, <65 3 points and <75 2 points. LV Hypertrophy was assigned 2 points, Type 1 Diabetes 3 points and Type 2 diabetes 1 point, LV Dysfunction was assigned 2 points. This scoring system had excellent discriminative ability (derivation AUROC 0.80, validation AUROC 0.81). A score of 8 or more had a specificity of 97% and a positive likelihood ratio of 9.4 for non-diagnostic ESE. A score of <8 had a negative predictive value of 86%. Patients with Non-Diagnostic ESE had longer time to transplantation (695 vs 636 days), and more investigations performed (43.05% vs 17.94%, p<0.001) and less likely to receive a renal transplant (OR 0.47, p<0.001). Conclusion Non diagnostic ESE in pre-transplant assessment causes significant delays to patient care. A simple clinical scoring system may identify patients likely to have non-diagnostic ESE and can be utilised to identify patients who may benefit from alternative risk-stratification investigations.

White Smoke in the Intermediate risk PE debate, could POPE be the solution?

Abstract Introduction In spite of a multitude of innovative therapeutic strategies emerging in the past decade, acute pulmonary embolism (PE) remains a potentially fatal disease. However, mortality varies significantly according to risk stratification, substantiating current recommendations to use reperfusion therapy only in hemodynamically unstable patients. There is an ongoing debate around whether hemodynamically stable patients showing signs of right ventricle dysfunction and myocardial injury (intermediate-high risk acute PE) may also benefit from reperfusion therapy. While the majority of patients in this category typically improve with anticoagulation alone, up to 10% may experience clinical deterioration. Aim We aimed to characterize the clinical in-hospital progression of patients with intermediate-risk PE and to evaluate the discriminative efficacy of PE scores in anticipating unfavorable clinical outcomes. Methods We conducted a retrospective observational study in patients admitted with the diagnosis of PE classified as intermediate risk in a tertiary center. Clinical, laboratory and ECG data were obtained. The BOVA score, the News score and the Pope score were evaluated. A composite endpoint of in-hospital mortality and use of reperfusion therapy due to hemodynamic deterioration was defined. Predictive abilities of these three scores were compared using area under the receiver operating characteristics (AUC-ROC) curve. Results From January 2019 to December 2020, a total of 166 patients were admitted with acute PE, classified as intermediate-risk, in our center. The mean age 71.9 ± 16.2 years and 57.2% were female. 71.1% of the pts showed bilateral PE in chest CT. The composite endpoint of in-hospital mortality and reperfusion therapy due to hemodynamic deterioration occurred in 9.8% of the pts. The median BOVA score was 3 (2-4), the median POPE score was 1 (0-2) and the median News score was 4 (2-6). The ROC curve analysis showed a significant higher discriminative power of POPE score compared to the other scores (AUC 0.856 95%CI 0.79–.92, p<0.001) (Fig1). Compared to patients with a POPE score 1, the composite endpoint was 4,6 times higher in patients with higher scores (OR 4.6 95%CI 1.63- 6914, p=0.029) (Fig2). Conclusion Despite classification as intermediate risk, these patients displayed a non-negligible in-hospital mortality rate. Close clinical monitoring is warranted in this group of pts and there is pressing need for an early identification of those who might benefit from more than just anticoagulation. In our population POPE score showed to be a simple and accurate tool in predicting those with an unfavorable evolution that could have benefitted from more advanced intervention.