Retrospective Dataset Research Articles

Developing new quantitative CT image markers to predict prognosis of acute ischemic stroke patients.

Endovascular mechanical thrombectomy (EMT) is an effective method to treat acute ischemic stroke (AIS) patients due to large vessel occlusion (LVO). However, stratifying AIS patients who can and cannot benefit from EMT remains a clinical challenge. To develop a new quantitative image marker computed from pre-intervention computed tomography perfusion (CTP) images and evaluate its feasibility to predict clinical outcome among AIS patients undergoing EMT after diagnosis of LVO. A retrospective dataset of 31 AIS patients with pre-intervention CTP images is assembled. A computer-aided detection (CAD) scheme is developed to pre-process CTP images of different scanning series for each study case, perform image segmentation, quantify contrast-enhanced blood volumes in bilateral cerebral hemispheres, and compute features related to asymmetrical cerebral blood flow patterns based on the cumulative cerebral blood flow curves of two hemispheres. Next, image markers based on a single optimal feature and machine learning (ML) models fused with multi-features are developed and tested to classify AIS cases into two classes of good and poor prognosis based on the Modified Rankin Scale. Performance of image markers is evaluated using the area under the ROC curve (AUC) and accuracy computed from the confusion matrix. The ML model using the neuroimaging features computed from the slopes of the subtracted cumulative blood flow curves between two cerebral hemispheres yields classification performance of AUC = 0.878±0.077 with an overall accuracy of 90.3%. This study demonstrates feasibility of developing a new quantitative imaging method and marker to predict AIS patients' prognosis in the hyperacute stage, which can help clinicians optimally treat and manage AIS patients.

A deep learning masked segmentation alternative to manual segmentation in biparametric MRI prostate cancer radiomics

ObjectivesTo determine the value of a deep learning masked (DLM) auto-fixed volume of interest (VOI) segmentation method as an alternative to manual segmentation for radiomics-based diagnosis of clinically significant (CS) prostate cancer (PCa) on biparametric magnetic resonance imaging (bpMRI).Materials and methodsThis study included a retrospective multi-center dataset of 524 PCa lesions (of which 204 are CS PCa) on bpMRI. All lesions were both semi-automatically segmented with a DLM auto-fixed VOI method (averaging < 10 s per lesion) and manually segmented by an expert uroradiologist (averaging 5 min per lesion). The DLM auto-fixed VOI method uses a spherical VOI (with its center at the location of the lowest apparent diffusion coefficient of the prostate lesion as indicated with a single mouse click) from which non-prostate voxels are removed using a deep learning–based prostate segmentation algorithm. Thirteen different DLM auto-fixed VOI diameters (ranging from 6 to 30 mm) were explored. Extracted radiomics data were split into training and test sets (4:1 ratio). Performance was assessed with receiver operating characteristic (ROC) analysis.ResultsIn the test set, the area under the ROC curve (AUCs) of the DLM auto-fixed VOI method with a VOI diameter of 18 mm (0.76 [95% CI: 0.66–0.85]) was significantly higher (p = 0.0198) than that of the manual segmentation method (0.62 [95% CI: 0.52–0.73]).ConclusionsA DLM auto-fixed VOI segmentation can provide a potentially more accurate radiomics diagnosis of CS PCa than expert manual segmentation while also reducing expert time investment by more than 97%.Key Points• Compared to traditional expert-based segmentation, a deep learning mask (DLM) auto-fixed VOI placement is more accurate at detecting CS PCa.• Compared to traditional expert-based segmentation, a DLM auto-fixed VOI placement is faster and can result in a 97% time reduction.• Applying deep learning to an auto-fixed VOI radiomics approach can be valuable.

Oral Ondansetron Reduces IV Fluid Use in the Emergency Department

Oral Ondansetron Reduces IV Fluid Use in the Emergency Department

A Computed Tomography Reconstruction Software To Model Ventricular Assist Device Implantation Position

<h3>Introduction</h3> As ventricular assist device (VAD) technology continues to evolve, the target population for left and right sided support continues to grow. Therefore the degree of variability in cardiac anatomy and geometry in potential VAD recipients is likely to increase. Studying the cardiac geometry in relation to HeartMate 3 or HVAD device dimensions can help preoperative evaluation for ideal cannulation placement, and thereby potentially improve long term outcomes. Aims: Our project aims to develop a program that virtually models a VAD into the three-dimensional-rendering of the myocardium and chest using computed tomography images. <h3>Methods and Results</h3> We created a Python code process, a 4D retrospective cardiac CT dataset and then integrated a high definition 3D VAD model. Rather than creating a full-fledged viewer and analytics tool, which would be too costly and time-intensive, we created a tool that can be used to complement the capabilities of professional imaging processing software such as 3mensio or Visage (Figure 1 Demonstrates the program graphical user interface). The user can adjust the viewer to better isolate the heart prior to simulated VAD implant placement. The DICOM directories produced by our program can be opened directly in the software already used in daily clinical imaging to allow for more in-depth analysis of desired implant placement and cannula positioning, which can aid in pre-operative and intra-operative planning. <h3>Conclusions</h3> The next steps of the project will be to enable VAD implant model to be propagated across the 3D scan time series, increase computational efficiency and view clarity, as well as output the resulting DICOM directory so it can be opened in Visage or a similar tool. In a future version of the project, we will include an automatic placement feature and fluid dynamic analysis to validate the proposed implant location.

Machine learning approaches for optimal prediction of liver fibrosis cruelty

&#x0D; &#x0D; &#x0D; Noninvasive assessment of severity of liver fibrosis is crucial for understanding histology and making decisions on antiviral treatment for chronic HBV in view of the associated risks of biopsy. aimed to develop a computer-assisted assessment system for the evaluation of liver disease severity by using machine leaning classifier based on physical-layer with serum markers. The retrospective data set, including 920 patients, was used to establish Decision Tree classifier (DTC), Random Forest Classifier (RFC), Logistic Regression Classifier (LRC), and Support Vector Classifier (SVC) for liver fibrosis severity assessment. Training and testing samples account for 50% of the data set, respectively. The best indicator combinations were selected in random combinations of 24 indicators including 67 108 760 group indicators by four different machine learning classier. The resulting classifiers prospectively tested in 50% testing patients, and the sensitivity, septicity, overall accuracy, and receiver operating characteristics (ROC) were used to compare four classifiers to existed 19 models. Results show that the RFC-based classifier system, with 9 indicators, is feasible to assess severity for liver fibrosis with diagnostic accuracy (greater than 0.83) superior to existing 19 models. Additional studies based on a large data set with full serum markers and imaging information are necessary to enhance diagnostic accuracy and to expand clinical application.&#x0D; &#x0D; &#x0D;

Applying Machine Learning Methods and Models to Explore the Structure of Traffic Accident Data

The problem of reducing the increasing number of road traffic accidents has become more relevant in recent years. According to the United Nations plan this number has to be halved by 2030. A very effective way to handle it is to apply the machine learning paradigm to retrospective road traffic accident datasets. This case study applies machine learning techniques to form typical “portraits” of drivers violating road traffic rules by clustering available data into seven homogeneous groups. The obtained results can be used in forming effective marketing campaigns for different target groups. Another relevant problem under consideration is to use available retrospective statistics on mechanical road traffic accidents without victims to estimate the probable type of road traffic accident for the driver taking into account her/his personal features (such as social characteristics, typical road traffic rule violations, driving experience, and age group). For this purpose several machine learning models were applied and the results were discussed.

Developing an Intelligent System for Diagnosis of COVID-19 Based on Artificial Neural Network

An outbreak of atypical pneumonia termed coronavirus disease 2019 (COVID-19) has spread worldwide since the beginning of 2020. It poses a significant threat to the global health and the economy. Physicians face ambiguity in their decision-making for COVID-19 diagnosis and treatment. In this respect, designing an intelligent system for early diagnosis of the disease is critical for mitigating virus spread and resource optimization. This study aimed to establish an artificial neural network (ANNs)-based clinical model to diagnose COVID-19. The retrospective dataset used in this study consisted of 400 COVID-19 case records (250 positives vs. 150 negatives) and 18 columns for the diagnostic features. The backpropagation technique was used to train a neural network. After designing multiple neural network configurations, the area under the receiver-operating characteristic curve (AUC), accuracy, sensitivity, and specificity values were calculated to measure the model performance. The two nested loops architecture of 9-10-15-2 (10 and 15 neurons used in layer one and layer two, respectively) with the ROC of 98.2%, sensitivity of 96.4%, specificity of 90.6%, and accuracy of 94 % were introduced as the best configuration model for COVID-19 diagnosis. ANN is valuable as a decision-support tool for clinicians to improve the COVID-19 diagnosis. It is promising to implement the ANN model to improve the accuracy and speed of the COVID-19 diagnosis for timely screening, treatment, and careful monitoring. Further studies are warranted for verifying and improving the current ANN model.

Improving Surgical Triage in Spine Clinic: Predicting Likelihood of Surgery Using Machine Learning

Correctly triaging patients to a surgeon or a nonoperative provider is an important part of the referral process. Clinics typically triage new patients based on simple intake questions. This is time-consuming and does not incorporate objective data. Our goal was to use machine learning to more accurately screen surgical candidates seen in a spine clinic. Using questionnaire data and magnetic resonance imaging reports, a set of artificial neural networks was trained to predict whether a patient would be recommended for spine surgery. Questionnaire responses included demographics, chief complaint, and pain characteristics. The primary end point was the surgeon's determination of whether a patient was an operative candidate. Model accuracy in predicting this end point was assessed using a separate subset of patients. The retrospective dataset included 1663 patients in cervical and lumbar cohorts. Questionnaire data were available for all participants, and magnetic resonance imaging reports were available for 242 patients. Within 6 months of initial evaluation, 717 (43.1%) patients were deemed surgical candidates by the surgeon. Our models predicted surgeons' recommendations with area under the curve scores of 0.686 for lumbar (positive predictive value 66%, negative predictive value 80%) and 0.821 for cervical (positive predictive value 83%, negative predictive value 85%) patients. Our models used patient data to accurately predict whether patients will receive a surgical recommendation. The high negative predictive value demonstrates that this approach can reduce the burden of nonsurgical patients in surgery clinic without losing many surgical candidates. This could reduce unnecessary visits for patients, increase the proportion of operative candidates seen by surgeons, and improve quality of patient care.

AI and Clinical Decision Making: The Limitations and Risks of Computational Reductionism in Bowel Cancer Screening

Advances in artificial intelligence in healthcare are frequently promoted as ‘solutions’ to improve the accuracy, safety, and quality of clinical decisions, treatments, and care. Despite some diagnostic success, however, AI systems rely on forms of reductive reasoning and computational determinism that embed problematic assumptions about clinical decision-making and clinical practice. Clinician autonomy, experience, and judgement are reduced to inputs and outputs framed as binary or multi-class classification problems benchmarked against a clinician’s capacity to identify or predict disease states. This paper examines this reductive reasoning in AI systems for colorectal cancer (CRC) to highlight their limitations and risks: (1) in AI systems themselves due to inherent biases in (a) retrospective training datasets and (b) embedded assumptions in underlying AI architectures and algorithms; (2) in the problematic and limited evaluations being conducted on AI systems prior to system integration in clinical practice; and (3) in marginalising socio-technical factors in the context-dependent interactions between clinicians, their patients, and the broader health system. The paper argues that to optimise benefits from AI systems and to avoid negative unintended consequences for clinical decision-making and patient care, there is a need for more nuanced and balanced approaches to AI system deployment and evaluation in CRC.

Evaluating the impact of social determinants, conditional cash transfers and primary health care on HIV/AIDS: Study protocol of a retrospective and forecasting approach based on the data integration with a cohort of 100 million Brazilians.

BackgroundDespite the great progress made over the last decades, stronger structural interventions are needed to end the HIV/AIDS pandemic in Low and Middle-Income Countries (LMIC). Brazil is one of the largest and data-richest LMIC, with rapidly changing socioeconomic characteristics and an important HIV/AIDS burden. Over the last two decades Brazil has also implemented the world’s largest Conditional Cash Transfer programs, the Bolsa Familia Program (BFP), and one of the most consolidated Primary Health Care (PHC) interventions, the Family Health Strategy (FHS).ObjectiveWe will evaluate the effects of socioeconomic determinants, BFP exposure and FHS coverage on HIV/AIDS incidence, treatment adherence, hospitalizations, case fatality, and mortality using unprecedently large aggregate and individual-level longitudinal data. Moreover, we will integrate the retrospective datasets and estimated parameters with comprehensive forecasting models to project HIV/AIDS incidence, prevalence and mortality scenarios up to 2030 according to future socioeconomic conditions and alternative policy implementations.Methods and analysisWe will combine individual-level data from all national HIV/AIDS registries with large-scale databases, including the “100 Million Brazilian Cohort”, over a 19-year period (2000–2018). Several approaches will be used for the retrospective quasi-experimental impact evaluations, such as Regression Discontinuity Design (RDD), Random Administrative Delays (RAD) and Propensity Score Matching (PSM), combined with multivariable Poisson regressions for cohort analyses. Moreover, we will explore in depth lagged and long-term effects of changes in living conditions and in exposures to BFP and FHS. We will also investigate the effects of the interventions in a wide range of subpopulations. Finally, we will integrate such retrospective analyses with microsimulation, compartmental and agent-based models to forecast future HIV/AIDS scenarios.ConclusionThe unprecedented datasets, analyzed through state-of-the-art quasi-experimental methods and innovative mathematical models will provide essential evidences to the understanding and control of HIV/AIDS epidemic in LMICs such as Brazil.

Translation of Prognostic and Pharmacodynamic Biomarkers from Trial to Non-trial Patients with Metastatic Castration-resistant Prostate Cancer Treated with Docetaxel

AimsWe conducted a pooled analysis of four randomised controlled trials and a non-trial retrospective dataset to study the changes in serum prostate-specific antigen (PSA) concentrations during treatment and its impact on survival in men treated with docetaxel for metastatic castration-resistant prostate cancer. We also compared the outcomes and pre-treatment prognostic factors between trial and non-trial patients. Materials and methodsData were obtained from four randomised controlled trials and a non-trial cohort from a tertiary cancer centre. The PSA kinetics covariates chosen were absolute value (PSAT), best percentage change (BPCH) and tumour growth rate (K). The association between the covariates collected and overall survival was assessed within a Cox proportional hazards model. How well a covariate captured the difference between trial and non-trial patients was assessed by reporting on models with or without trial status as a covariate. ResultsWe reviewed individual datasets of 2282 patients. The median overall survival for trial patients was 20.4 (95% confidence interval 19.6–22.2) months and for the non-trial cohort was 12.4 (10.7–14.7) months (P < 0.001). Of the pre-treatment factors, we found that only lactate dehydrogenase fully captured the difference in prognosis between the trial and non-trial cohorts. All PSA kinetic metrics appeared to be prognostic in both the trial and non-trial patients. However, the effect size was reduced in non-trial versus trial patients (interaction P < 0.001). Of the time-dependent covariates, we found that BPCH best captured the difference between trial and non-trial patient prognosis. ConclusionsThe analysis presented here highlights how data from open-source trial databases can be combined with emerging clinical practice databases to assess differences between trial versus non-trial patients for particular treatments. These results highlight the importance of developing prognostic models using both pre-treatment and time-dependent biomarkers of new treatments.

Applying a radiomics-based CAD scheme to classify between malignant and benign pancreatic tumors using CT images.

Pancreatic cancer is one of the most aggressive cancers with approximate 10% five-year survival rate. To reduce mortality rate, accurate detection and diagnose of suspicious pancreatic tumors at an early stage plays an important role. To develop and test a new radiomics-based computer-aided diagnosis (CAD) scheme of computed tomography (CT) images to detect and classify suspicious pancreatic tumors. A retrospective dataset consisting of 77 patients who had suspicious pancreatic tumors detected on CT images was assembled in which 33 tumors are malignant. A CAD scheme was developed using the following 5 steps namely, (1) apply an image pre-processing algorithm to filter and reduce image noise, (2) use a deep learning model to detect and segment pancreas region, (3) apply a modified region growing algorithm to segment tumor region, (4) compute and select optimal radiomics features, and (5) train and test a support vector machine (SVM) model to classify the detected pancreatic tumor using a leave-one-case-out cross-validation method. By using the area under receiver operating characteristic (ROC) curve (AUC) as an evaluation index, SVM model yields AUC = 0.750 with 95% confidence interval [0.624, 0.885] to classify pancreatic tumors. Study results indicate that radiomics features computed from CT images contain useful information associated with risk of tumor malignancy. This study also built a foundation to support further effort to develop and optimize CAD schemes with more advanced image processing and machine learning methods to more accurately and robustly detect and classify pancreatic tumors in future.

Performance Comparison of Infection Prediction Scores in a South African Neonatal Unit: A Retrospective Case-Control Study.

Background and objectivesInfection prediction scores are useful ancillary tests in determining the likelihood of neonatal hospital-acquired infection (HAI), particularly in very low birth weight (VLBW; <1,500 g) infants who are most vulnerable to HAI and have high antibiotic utilization rates. None of the existing infection prediction scores were developed for or evaluated in South African VLBW neonates.MethodsWe identified existing infection prediction scores through literature searches and assessed each score for suitability and feasibility of use in resource-limited settings. Performance of suitable scores were compared using a retrospective dataset of VLBW infants (2016–2017) from a tertiary hospital neonatal unit in Cape Town, South Africa. Sensitivity, specificity, predictive values, and likelihood ratios were calculated for each score.ResultsEleven infection prediction scores were identified, but only five were suitable for use in resource-limited settings (NOSEP1, Singh, Rosenberg, and Bekhof scores). The five selected scores were evaluated using data from 841 episodes of HAI in 659 VLBW infants. The sensitivity for the scores ranged between 3% (NOSEP1 ≥14; proven and presumed infection), to a maximum of 74% (Singh score ≥1; proven infection). The specificity of these scores ranged from 31% (Singh score ≥1; proven and presumed infection) to 100% (NOSEP1 ≥11 and ≥14, NOSEP-NEW-1 ≥11; proven and presumed infection).ConclusionExisting infection prediction scores did not achieve comparable predictive performance in South African VLBW infants and should therefore only be used as an adjunct to clinical judgment in antimicrobial decision making. Future studies should develop infection prediction scores that have high diagnostic accuracy and are feasible to implement in resource-limited neonatal units.

An Online Pre-procedural Nomogram for the Prediction of Contrast-Associated Acute Kidney Injury in Patients Undergoing Coronary Angiography.

BackgroundIdentifying high-risk patients for contrast-associated acute kidney injury (CA-AKI) helps to take early preventive interventions. The current study aimed to establish and validate an online pre-procedural nomogram for CA-AKI in patients undergoing coronary angiography (CAG).MethodsIn this retrospective dataset, 4,295 patients undergoing CAG were enrolled and randomized into the training or testing dataset with a split ratio of 8:2. Optimal predictors for CA-AKI were determined by Least Absolute Shrinkage and Selection Operator (LASSO) and Random Forest (RF) algorithm. Nomogram was developed and deployed online. The discrimination and accuracy of the nomogram were evaluated by receiver operating characteristic (ROC) and calibration analysis, respectively. Clinical usefulness was estimated by decision curve analysis (DCA) and clinical impact curve (CIC).ResultsA total of 755 patients (17.1%) was diagnosed with CA-AKI. 7 pre-procedural predictors were identified and integrated into the nomogram, including age, gender, hemoglobin, N-terminal of the prohormone brain natriuretic peptide, neutrophil-to-lymphocyte ratio, cardiac troponin I, and loop diuretics use. The ROC analyses showed that the nomogram had a good discrimination performance for CA-AKI in the training dataset (area under the curve, AUC = 0.766, 95%CI [0.737 to 0.794]) and testing dataset (AUC = 0.737, 95%CI [0.693 to 0.780]). The nomogram was also well-calibrated in both the training dataset (P = 0.965) and the testing dataset (P = 0.789). Good clinical usefulness was identified by DCA and CIC. Finally, this model was deployed in a web server for public use (https://duanbin-li.shinyapps.io/DynNomapp/).ConclusionAn easy-to-use pre-procedural nomogram for predicting CA-AKI was established and validated in patients undergoing CAG, which was also deployed online.

Oropharyngeal Shedding of Gammaherpesvirus DNA by Cats, and Natural Infection of Salivary Epithelium.

Felis catus gammaherpesvirus-1 (FcaGHV1), a novel candidate oncogenic virus, infects cats worldwide. Whether the oropharynx is a site of virus shedding and persistence, and whether oronasal carcinomas harbor FcaGHV1 nucleic acid were investigated. In a prospective molecular epidemiological study, FcaGHV1 DNA was detected by cPCR in oropharyngeal swabs from 26/155 (16.8%) of cats. Oropharyngeal shedding was less frequently detected in kittens ≤3 months of age (5/94, 5.3%) than in older animals; >3 months to ≤1 year: 8/26, 30.8%, (p = 0.001, OR 7.91, 95% CI (2.320, 26.979)); >1 year to ≤6 years: 10/20, 50%, (p < 0.001, OR 17.8 95% CI (5.065, 62.557)); >6 years: 3/15, 33% (p = 0.078). Provenance (shelter-owned/privately owned) was not associated with shedding. In situ hybridization (ISH) identified FcaGHV1-infected cells in salivary glandular epithelium but not in other oronasal tissues from two of three cats shedding viral DNA in the oropharynx. In a retrospective dataset of 11 oronasopharyngeal carcinomas, a single tumor tested positive for FcaGHV1 DNA by ISH, a papillary carcinoma, where scattered neoplastic cells showed discrete nuclear hybridization. These data support the oronasopharynx as a site of FcaGHV1 shedding, particularly after maternal antibodies are expected to decline. The salivary epithelium is identified as a potential site of FcaGHV1 persistence. No evidence supporting a role for FcaGHV1 in feline oronasal carcinomas was found in the examined tumours.

Toward automated segmentation for acute ischemic stroke using non-contrast computed tomography.

Non-contrast computed tomography (NCCT) is a first-line imaging technique for determining treatment options for acute ischemic stroke (AIS). However, its poor contrast and signal-to-noise ratio limit the diagnosis accuracy for radiologists, and automated AIS lesion segmentation using NCCT also remains a challenge. In this paper, we propose R2U-RNet, a novel model for AIS lesion segmentation using NCCT. We used an in-house retrospective NCCT dataset with 261 AIS patients with manual lesion segmentation using follow-up diffusion-weighted images. R2U-RNet is based on an R2U-Net backbone with a novel residual refinement unit. Each input image contains two image channels from separate preprocessing procedures. The proposed model incorporates multiscale focal loss to mitigate the class imbalance problem and to leverage the importance of different levels of details. A proposed noisy-label training scheme is utilized to account for uncertainties in the manual annotations. The proposed model outperformed several iconic segmentation models in AIS lesion segmentation using NCCT, and our ablation study demonstrated the efficacy of the proposed model. Statistical analysis of segmentation performance revealed significant effects of regional stroke occurrence and side of the stroke, suggesting the importance of region-specific information for automated segmentation, and the potential influence of the hemispheric difference in clinical data. This study demonstrated the potentials of R2U-RNet model for automated NCCT AIS lesion segmentation. The proposed model can serve as a tool for accelerating AIS diagnoses and improving the treatment quality of AIS patients.

Cross-institutional outcome prediction for head and neck cancer patients using self-attention neural networks

In radiation oncology, predicting patient risk stratification allows specialization of therapy intensification as well as selecting between systemic and regional treatments, all of which helps to improve patient outcome and quality of life. Deep learning offers an advantage over traditional radiomics for medical image processing by learning salient features from training data originating from multiple datasets. However, while their large capacity allows to combine high-level medical imaging data for outcome prediction, they lack generalization to be used across institutions. In this work, a pseudo-volumetric convolutional neural network with a deep preprocessor module and self-attention (PreSANet) is proposed for the prediction of distant metastasis, locoregional recurrence, and overall survival occurrence probabilities within the 10 year follow-up time frame for head and neck cancer patients with squamous cell carcinoma. The model is capable of processing multi-modal inputs of variable scan length, as well as integrating patient data in the prediction model. These proposed architectural features and additional modalities all serve to extract additional information from the available data when availability to additional samples is limited. This model was trained on the public Cancer Imaging Archive Head–Neck-PET–CT dataset consisting of 298 patients undergoing curative radio/chemo-radiotherapy and acquired from 4 different institutions. The model was further validated on an internal retrospective dataset with 371 patients acquired from one of the institutions in the training dataset. An extensive set of ablation experiments were performed to test the utility of the proposed model characteristics, achieving an AUROC of 80%, 80% and 82% for DM, LR and OS respectively on the public TCIA Head–Neck-PET–CT dataset. External validation was performed on a retrospective dataset with 371 patients, achieving 69% AUROC in all outcomes. To test for model generalization across sites, a validation scheme consisting of single site-holdout and cross-validation combining both datasets was used. The mean accuracy across 4 institutions obtained was 72%, 70% and 71% for DM, LR and OS respectively. The proposed model demonstrates an effective method for tumor outcome prediction for multi-site, multi-modal combining both volumetric data and structured patient clinical data.

Use of deep learning in the MRI diagnosis of Chiari malformation type I

PurposeTo train deep learning convolutional neural network (CNN) models for classification of clinically significant Chiari malformation type I (CM1) on MRI to assist clinicians in diagnosis and decision making.MethodsA retrospective MRI dataset of patients diagnosed with CM1 and healthy individuals with normal brain MRIs from the period January 2010 to May 2020 was used to train ResNet50 and VGG19 CNN models to automatically classify images as CM1 or normal. A total of 101 patients diagnosed with CM1 requiring surgery and 111 patients with normal brain MRIs were included (median age 30 with an interquartile range of 23–43; 81 women with CM1). Isotropic volume transformation, image cropping, skull stripping, and data augmentation were employed to optimize model accuracy. K-fold cross validation was used to calculate sensitivity, specificity, and the area under receiver operating characteristic curve (AUC) for model evaluation.ResultsThe VGG19 model with data augmentation achieved a sensitivity of 97.1% and a specificity of 97.4% with an AUC of 0.99. The ResNet50 model achieved a sensitivity of 94.0% and a specificity of 94.4% with an AUC of 0.98.ConclusionsVGG19 and ResNet50 CNN models can be trained to automatically detect clinically significant CM1 on MRI with a high sensitivity and specificity. These models have the potential to be developed into clinical support tools in diagnosing CM1.

Development and validation of health system performance measures for opioid use disorder in British Columbia, Canada

BackgroundPerformance measurement provides an evidence-based means to inform development of interventions to improve the quality of care for people who use opioids. We aimed to develop and assess the predictive validity of health system performance measures for opioid use disorder (OUD) in British Columbia (BC), Canada. MethodsPerformance measures were generated using retrospective population-level administrative datasets (both provincial and regional) and publicly-reported retrospective data according to four domains (care engagement, clinical guideline compliance, integration, and healthcare utilization). The adjusted odds ratio was estimated via generalized linear mixed models to determine predictive validity for all-cause hospitalization or mortality within 6 months of measurement. FindingsA total of 102 performance measures were constructed. We identified 55,470 diagnosed PWOUD, and 39,456 ever engaged in opioid agonist treatment (OAT). We found divergent rates of treatment for concurrent conditions (7.4% for alcohol use disorder to 80.1% for HIV/AIDS), low levels of linkage to OAT and other outpatient care following acute care, and increasing levels of service provision, including increases in OAT prescribers and pharmacies, naloxone kit distribution and overdose prevention site visitation. Our analyses on the predictive validity measures largely supported a priori hypotheses on the direction of effect on the outcome. ConclusionsWe identified a range of priorities to improve the quality of care for PWOUD, with critical gaps in linkage to care through acute care settings and long-term engagement in OAT. The proposed measures can be derived for geographic and clinical subgroups and updated over time, providing a basis to monitor and evaluate efforts to address the public health burden of OUD.

Sudden Sensorineural Hearing Loss May Increase the Risk of Retinal Vein Occlusion: A Nationwide Cohort Study.

Sudden sensorineural hearing loss (SSNHL) is thought to be a vascular disease. Retinal vein occlusion (RVO) is an also common ophthalmic vascular disease. Thus, we investigated the potential relationship between these using a retrospective nationwide cohort dataset. We compared 49,584 subjects in the SSNHL and the comparison (non-SSNHL) groups using patients randomly selected via propensity-score matching. We calculated the incidence, survival rate, and hazards ratio (HR) using log-rank test, and Cox proportional hazards regression models. This study examined a total of 375,490.4 person-years in the SSNHL group and 373,698.2 person-years in the comparison group. We found that 673 patients in the SSNHL group (1.8 cases per 1000 person-years) and 592 in the comparison group (1.6 cases per 1000 person-years) developed RVO during the 8-year follow-up period. The adjusted HR of RVO was 1.13 (95% confidence interval [CI] 1.01–1.26). The adjusted HR of developing RVO in SSNHL was the greatest in elderly patients (adjusted HR, 1.21; 95% CI, 1.01–1.46) and male patients (adjusted HR, 1.18; 95% CI, 1.03–1.34). Our findings suggest that clinicians should remain vigilant of the possibility of RVO development in SSNHL patients, specifically elderly male patients.