Search

Introduction and objectives Optimal treatment of hepatocellular carcinoma (HCC) involving portal vein tumor thrombus (PVTT) remains controversial. Materials and methods A total of 627 HCC patients with PVTT after initial treatment with one of the following at Affiliated Tumor Hospital of Guangxi Medical University: liver resection (LR, n = 225), transarterial chemoembolization (TACE, n = 298) or sorafenib (n = 104) were recruited and randomly divided into the training cohort (n = 314) and internal validation cohort (n = 313). Survival analysis were repeated after stratifying patients by Cheng PVTT type. Results Resection led to significantly higher OS than the other two treatments among patients with type I or II PVTT. TACE worked significantly better than the other two treatments for patients with type III. All three treatments were associated with similar OS among patients with type IV. These findings were supported by the internal validation cohort. Conclusions Our results suggest that the optimal treatment for HCC involving PVTT depends on the type of PVTT. LR may be more appropriate for type I or II PVTT; TACE, for type III Sorafenib may be more appropriate than invasive treatments for patients with type IV PVTT.

BackgroundTo establish a machine-learning-derived nomogram based on radiomic features and clinical factors to predict post-surgical 2-year progression-free survival (PFS) in patients with lung adenocarcinoma.MethodsPatients with >2 years post-surgical prognosis results of lung adenocarcinoma were included in Hospital-1 for model training (n = 100) and internal validation (n = 50), and in Hospital-2 for external testing (n = 50). A total of 1,672 radiomic features were extracted from 3D segmented CT images. The Rad-score was established using random survival forest by accumulating and weighting the top-20 imaging features contributive to PFS. A nomogram for predicting PFS was established, which comprised the Rad-score and clinical factors highly relevant to PFS.ResultsIn the training, internal validation, and external test groups, 69/100 (69%), 37/50 (74%) and 36/50 (72%) patients were progression-free at two years, respectively. According to the Rad-score, the integral of area under the curve (iAUC) for discriminating high and low risk of progression was 0.92 (95%CI: 0.77-1.0), 0.70 (0.41-0.98) and 0.90 (0.65-1.0), respectively. The C-index of Rad-score was 0.781 and 0.860 in the training and external test groups, higher than 0.707 and 0.606 for TNM stage, respectively. The nomogram integrating Rad-score and clinical factors (lung nodule type, cM stage and histological type) achieved a C-index of 0.845 and 0.837 to predict 2-year PFS, respectively, significantly higher than by only radiomic features (all p < 0.01).ConclusionThe nomogram comprising CT-derived radiomic features and risk factors showed a high performance in predicting post-surgical 2-year PFS of patients with lung adenocarcinoma, which may help personalize the treatment decisions.

Three clinical prediction models were developed for nonrecovery in older adults seeking care for back pain; the model predicting persistent disability performed best (optimism-adjusted R2 46%, area under the curve 0.85)., Back pain is a leading cause of disability worldwide and is common in older adults. No clinical prediction models for poor long-term outcomes have been developed in older patients with back pain. This study aimed to develop and internally validate 3 clinical prediction models for nonrecovery in this population. A prospective cohort study in general practice was conducted (Back Complaints in the Elders, Netherlands), including 675 patients >55 years with a new episode of care for back pain. Three definitions of nonrecovery were used combining 6-month and 12-month follow-up data: (1) persistent back pain, (2) persistent disability, and (3) perceived nonrecovery. Sample size calculation resulted in a maximum of 14 candidate predictors that were selected from back pain prognostic literature and clinical experience. Multivariable logistic regression was used to develop the models (backward selection procedure). Models' performance was evaluated with explained variance (Nagelkerke's R2), calibration (Hosmer–Lemeshow test), and discrimination (area under the curve [AUC]) measures. The models were internally validated in 250 bootstrapped samples to correct for overoptimism. All 3 models displayed good overall performance during development and internal validation (ie, R2 > 30%; AUC > 0.77). The model predicting persistent disability performed best, showing good calibration, discrimination (AUC 0.86, 95% confidence interval 0.83-0.89; optimism-adjusted AUC 0.85), and explained variance (R2 49%, optimism-adjusted R2 46%). Common predictors in all models were: age, chronic duration, disability, a recent back pain episode, and patients' recovery expectations. Spinal morning stiffness and pain during spinal rotation were included in 2 of 3 models. These models should be externally validated before being used in a clinical primary care setting.

Background: in magnetic resonance imaging (MRI), automated detection of brain metastases with convolutional neural networks (CNN) represents an extraordinary challenge due to small lesions sometimes posing as brain vessels as well as other confounders. Literature reporting high false positive rates when using conventional contrast enhanced (CE) T1 sequences questions their usefulness in clinical routine. CE black blood (BB) sequences may overcome these limitations by suppressing contrast-enhanced structures, thus facilitating lesion detection. This study compared CNN performance in conventional CE T1 and BB sequences and tested for objective improvement of brain lesion detection. Methods: we included a subgroup of 127 consecutive patients, receiving both CE T1 and BB sequences, referred for MRI concerning metastatic spread to the brain. A pretrained CNN was retrained with a customized monolayer classifier using either T1 or BB scans of brain lesions. Results: CE T1 imaging-based training resulted in an internal validation accuracy of 85.5% vs. 92.3% in BB imaging (p < 0.01). In holdout validation analysis, T1 image-based prediction presented poor specificity and sensitivity with an AUC of 0.53 compared to 0.87 in BB-imaging-based prediction. Conclusions: detection of brain lesions with CNN, BB-MRI imaging represents a highly effective input type when compared to conventional CE T1-MRI imaging. Use of BB-MRI can overcome the current limitations for automated brain lesion detection and the objectively excellent performance of our CNN suggests routine usage of BB sequences for radiological analysis.

Background Canine leishmaniosis (CanL) can be appropriately managed following international recommendations. However, few studies have assessed the preferred protocols in real-life veterinary practice and whether these are in line with the guidelines. This survey aimed to investigate the current trends in the clinical management of CanL among veterinary practitioners in Portugal, taking into consideration different scenarios of infection/disease and the awareness of and application by veterinary practitioners of the current guidelines. Methods A questionnaire-based survey was conducted online using an electronic platform. The following topics were surveyed: (i) general characteristics of the responding veterinarian; (ii) the preferred protocols used for the diagnosis, treatment and prevention of CanL, considering different theoretical scenarios of infection/disease; and (iii) the responding veterinarian’s current knowledge and application of the existing guidelines on CanL. After internal validation, the survey was distributed online, for 2 months, via Portuguese social network veterinary groups. Data were collected for descriptive analysis. Results Eighty-six replies were obtained. Analysis of the results showed that the preferred diagnostic techniques varied widely according to the theoretical scenario of infection/disease. In general daily practice, serology testing (enzyme-linked immunosorbent assay [ELISA]) was the most used tool (67.4%). The preferred matrices used for PCR test were lymph nodes (62.3%) and/or bone marrow (59.0%). Regarding treatment, for subclinical infection/stage I CanL, 51.2% of the respondents did not prescribe any medical treatment, but 98.8% proceeded with both monitoring and preventive measures. Among those who prescribed a treatment (n = 42), most chose domperidone (47.6%). For the treament of stages IIa, IIb and III CanL, allopurinol/meglumine antimoniate (MA) was chosen by 69.8, 73.3 and 51.2% of respondents, respectively, followed by allopurinol/miltefosine (20.9, 19.8 and 38.4%, respectively). In contrast, dogs with stage IV CanL were mostly treated with allopurinol/miltefosine (48.8%) rather than with allopurinol/MA (23.3%). The use of repellents was the preferred preventive strategy (98.8%). About 93.0% of responders were aware of the existence of guidelines, and most of these veterinarians consulted the guidelines of the LeishVet group and the Canine Leishmaniosis Working Group; however, 31.3% reported that they did not follow any specific recommendations. Conclusions Of the veterinarians responding to the survey, most reported following international guidelines for the clinical management of CanL. While allopurinol/MA was the preferred therapeutic protocol for the treatment of stages II/III CanL, allopurinol/miltefosine was the first choice for the treatment of stage IV CanL, possibly due to the unpredictable effect of MA on renal function. This study contributes to a better understanding of the trends in practical approaches to the treatment of CanL in Portugal. Graphical Abstract

Clarifying whether physiological sleep measures predict mortality could inform risk screening; however, such investigations should account for complex and potentially non-linear relationships among health risk factors. We aimed to establish the predictive utility of polysomnography-assessed sleep measures for mortality using a novel Permutation Random Forest machine learning framework. Data collected from years 1995 – present are from the Sleep Heart Health Study (SHHS; n=5734) and the Wisconsin Sleep cohort study (WSCS; n=1015) and include initial assessments of sleep and heath and up to 15 years of follow-up for all-cause mortality. We applied Permutation Random Forest models to quantify the predictive abilities of 24 measures grouped into 5 domains: polysomnography-assessed sleep (4 measures), self-reported sleep (3), health (8), health behaviors (4), and sociodemographic factors (5). Ten-fold repeated internal validation (WSCS and SHHS combined) and external validation (training in SHHS; testing in WSCS) were used to compute unbiased variable importance metrics and associated p-values. We observed that health, sociodemographic factors, and polysomnography-assessed sleep domains predicted mortality using both External Validation and repeated Internal Validation. Polysomnography-assessed sleep efficiency and the percentage of sleep time with oxygen saturation < 90% (T90) were among the most predictive individual measures. Multivariable Cox regression also revealed the polysomnography-assessed sleep domain to be predictive, with very low sleep efficiency and high hypoxemia conferring the highest risk. These findings, coupled with the emergence of new low-burden technologies for objectively assessing sleep and overnight oxygen saturation, suggest that consideration of physiological sleep measures may improve risk screening.

Hydrogen breath tests are a well-established method to help diagnose functional intestinal disorders such as carbohydrate malabsorption or small intestinal bacterial overgrowth. In this work we apply unsupervised machine learning techniques to analyze hydrogen breath test datasets. We propose a method that uses 26 internal cluster validation measures to determine a suitable number of clusters. In an induced external validation step we use a predefined categorization proposed by a medical expert. The results indicate that the majority of the considered internal validation indexes was not able to produce a reasonable clustering. Considering a predefined categorization performed by a medical expert, a novel shape-based method obtained the highest external validation measure in terms of adjusted rand index. The predefined clusterings constitute the basis of a supervised machine learning step that is part of our ongoing research.

Several precision livestock farming (PLF) technologies, conceived for optimizing farming processes, are developed to detect the physical and behavioral changes of animals continuously and in real-time. The aim of this review was to explore the capacity of existing PLF technologies to contribute to the assessment of pig welfare. In a web search for commercially available PLF for pigs, 83 technologies were identified. A literature search was conducted, following systematic review guidelines (PRISMA), to identify studies on the validation of sensor technologies for assessing animal-based welfare indicators. Two validation levels were defined: internal (evaluation during system building within the same population that were used for system building) and external (evaluation on a different population than during system building). From 2,463 articles found, 111 were selected, which validated some PLF that could be applied to the assessment of animal-based welfare indicators of pigs (7% classified as external, and 93% as internal validation). From our list of commercially available PLF technologies, only 5% had been externally validated. The more often validated technologies were vision-based solutions (n = 45), followed by load-cells (n = 28; feeders and drinkers, force plates and scales), accelerometers (n = 14) and microphones (n = 14), thermal cameras (n = 10), photoelectric sensors (n = 5), radio-frequency identification (RFID) for tracking (n = 2), infrared thermometers (n = 1), and pyrometer (n = 1). Externally validated technologies were photoelectric sensors (n = 2), thermal cameras (n = 2), microphone (n = 1), load-cells (n = 1), RFID (n = 1), and pyrometer (n = 1). Measured traits included activity and posture-related behavior, feeding and drinking, other behavior, physical condition, and health. In conclusion, existing PLF technologies are potential tools for on-farm animal welfare assessment in pig production. However, validation studies are lacking for an important percentage of market available tools, and in particular research and development need to focus on identifying the feature candidates of the measures (e.g., deviations from diurnal pattern, threshold levels) that are valid signals of either negative or positive animal welfare. An important gap identified are the lack of technologies to assess affective states (both positive and negative states).

Made available in DSpace on 2021-06-25T10:30:44Z (GMT). No. of bitstreams: 0 Previous issue date: 2021-05-01 Background Predicting waiting time for a deceased donor kidney transplant can help patients and clinicians to discuss management and contribute to a more efficient use of resources. This study aimed at developing a predictor model to estimate time on a kidney transplant waiting list using a machine learning approach. Methods A retrospective cohort study including data of patients registered, between January 1, 2000 and December 31, 2017, in the waiting list of São Paulo State Organ Allocation System (SP-OAS) /Brazil. Data were randomly divided into two groups: 75% for training and 25% for testing. A Cox regression model was fitted with deceased donor transplant as the outcome. Sensitivity analyses were performed using different Cox models. Cox hazard ratios were used to develop the risk-prediction equations. Results Of 54,055 records retrieved, 48,153 registries were included in the final analysis. During the study period, approximately 1/3 of the patients were transplanted with a deceased donor. The major characteristics associated with changes in the likelihood of transplantation were age, subregion, cPRA, and frequency of HLA-DR, -B and -A. The model developed was able to predict waiting time with good agreement in internal validation (c-index = 0.70). Conclusion The kidney transplant waiting time calculator developed shows good predictive performance and provides information that may be valuable in assisting candidates and their providers. Moreover, it can significantly improve the use of economic resources and the management of patient care before transplant. Department of Internal Medicine UNESP Univ Estadual Paulista Transplant Unit–Santa Casa Juiz de Fora Kidney-Pancreas Transplantation Service of Leforte and Oswaldo Cruz Hospitals Transplant Unit Division Transplant Unit Division Liberdade Department of Internal Medicine UNESP Univ Estadual Paulista

BackgroundIschemic cardiovascular diseases (ICVD) risk predict models are valuable but limited by its requirement for multidimensional medical information including that from blood drawing. A convenient and affordable alternative is in demand.ObjectivesTo develop and validate a deep learning algorithm to predict 10-year ICVD risk using retinal fundus photographs in Chinese population.MethodsWe firstly labeled fundus photographs with natural logarithms of ICVD risk estimated by a previously validated 10-year Chinese ICVD risk prediction model for 390,947 adults randomly selected (95%) from a health checkup dataset. An algorithm using convolutional neural network was then developed to predict the estimated 10-year ICVD risk by fundus images. The algorithm was validated using both internal dataset (the other 5%) and external dataset from an independent source (sample size = 1,309). Adjusted R2and area under the receiver operating characteristic curve (AUC) were used to evaluate the goodness of fit.ResultsThe adjusted R2between natural logarithms of the predicted and calculated ICVD risks was 0.876 and 0.638 in the internal and external validations, respectively. For detecting ICVD risk ≥ 5% and ≥ 7.5%, the algorithm achieved an AUC of 0.971 (95% CI: 0.967–0.975) and 0.976 (95% CI: 0.973–0.980) in internal validation, and 0.859 (95% CI: 0.822–0.895) and 0.876 (95% CI: 0.816–0.837) in external validation.ConclusionsThe deep learning algorithm developed in the study using fundus photographs to predict 10-year ICVD risk in Chinese population had fairly good capability in predicting the risk and may have values to be widely promoted considering its advances in easy use and lower cost. Further studies with long term follow up are warranted.

Acute kidney injury is a common critical disease with a high mortality. The large number of indicators in AKI patients makes it difficult for clinicians to quickly and accurately determine the patient’s condition. This study used machine learning methods to filter key indicators and use key indicator data to achieve advance prediction of AKI so that a small number of indicators could be measured to reliably predict AKI and provide auxiliary decision support for clinical staff. Sequential forward selection based on feature importance calculated by XGBoost was used to screen out 17 key indicators. Three machine learning algorithms were used to make predictions, namely, logistic regression (LR), decision tree, and XGBoost. To verify the validity of the method, data were extracted from the MIMIC III database and the eICU-CRD database for 1,009 and 1,327 AKI patients, respectively. The MIMIC III database was used for internal validation, and the eICU-CRD database was used for external validation. For all three machine learning algorithms, the prediction performance from using only the key indicator dataset was very close to that from using the full dataset. The XGBoost algorithm performed the best, and LR was the next best. The decision tree performed the worst. The key indicator screening method proposed in this study can achieve a good predictive performance while streamlining the number of indicators.

Here we present an inexpensive, rapid, and robust RT-LAMP based SARS-CoV-2 detection method that is easily scalable, enabling point of care facilities and clinical labs to determine results from patients’ saliva directly in 30 minutes for less than $2 a sample. The method utilizes a novel combination of widely available reagents that can be prepared in bulk, plated and frozen and remain stable until samples are received. This innovation dramatically reduces preparation time, enabling high-throughput automation and testing with time to results (including setup) in less than one hour for 96 patient samples simultaneously when using a 384 well format. By utilizing a dual-reporter (phenol red pH indicator for end-point detection and SYTO-9 fluorescent dye for real-time), the assay also provides internal validation of results and redundancy in the event of an instrument malfunction.

BACKGROUND AND PURPOSE: Small intracranial aneurysms are being increasingly detected while the rupture risk is not well-understood. We aimed to develop rupture-risk models of small aneurysms by combining clinical, morphologic, and hemodynamic information based on machine learning techniques and to test the models in external validation datasets. MATERIALS AND METHODS: From January 2010 to December 2016, five hundred four consecutive patients with only small aneurysms (

The 6-min walk test (6MWT) is a simple method of identifying patients with a high risk of postoperative complications. In this study, we internally validated the previously obtained threshold value of 500 m in the 6MWT as differentiating populations with a high and a low risk of postoperative complications after a lobectomy. Between November 2011 and November 2016, 624 patients who underwent a lobectomy and performed the 6MWT preoperatively entered this study. We compared the complication rates of two groups of patients—those who walked more than and those who walked less than 500 m. The patients who did not reach the distance of 500 m in the 6MWT were older (70 vs. 63 years p &lt, 0.001), had worse pulmonary function tests (FEV1% 84 vs. 88 p = 0.041) and had a higher Charlson Comorbidity Index (p &lt, 0.001). The patients who had a worse result in the 6MWT had a higher complication rate (52% vs. 42% p = 0.019, OR: 1.501 95% CI: 1.066–2.114) and a longer median postoperative hospital stay (7 vs. 6 days p = 0.010). In a multivariate analysis, the result of the 6MWT and pack-years proved to independently influence the risk of postoperative complications. This internal validation study confirms that 500 m is a result of the 6MWT which differentiates patients with a higher risk of postoperative complications and a prolonged hospital stay after a lobectomy.

PURPOSE/OBJECTIVE(S) Radiotherapy is a standard palliative treatment for spine metastasis, but pain flares (PF) are a common acute toxicity. Prophylactic corticosteroids can reduce the rate of PF in patients receiving traditional palliative regimens, but less is known about prophylaxis for stereotactic body radiotherapy (SBRT). We aim to identify a subset of patients treated with spine SBRT with the highest rate of PF to optimize the use of prophylactic steroids. MATERIALS/METHODS From December 2007-August 2019, 469 patients received 680 spine SBRT treatments. After exclusion of benign histology, particle therapy, and patients with missing data, the final cohort included 424 patients with 610 treatments. We defined PF as acute worsening of pain at the treatment site requiring initiation of or higher-dose corticosteroids, opiates, and/or hospitalization. First, data was split into 70% training and 30% validation sets with comparable patient characteristics using a random number generator. Feature importance testing was conducted, and a correlation heatmap helped exclude correlated variables. Feature extraction involved selecting the variables used in the highest-fidelity recursive partitioning analysis (RPA) models. Each RPA model was trained, validated, and tested using only the training dataset. Selected variables were verified using the Fisher exact test and univariate t-test. RESULTS We identified 125 total PF (20%), treated with steroids (14%), opioids (40%), both (39%), or hospitalization (6%). Six variables were identified by RPA. Of these, 5 met significance on Fisher exact test with P 0 (OR 2.17), SINS > 6 (OR 2.98), and gross tumor volume > 8 cc (OR 2.80). One point was assigned for each variable. The low-risk (LR) group (score = 0, n = 159) had PF rates of 7.0% and 13.6% in the training and validation sets, respectively; the intermediate-risk (IR) group (score = 1, n = 150) had rates of 14.0% and 16.3%; and the high-risk (HR) group (score > 1, n = 301) had rates of 28.8% and 31.3%. A logistic model confirmed the increased pain flare rate in the HR group compared to LR and IR treatments combined (OR = 3.50, 95% CI: 2.06-5.92) or considered separately (OR = 5.41, 2.48-11.78). The concordance index was 0.66 (0.60-0.73) and 0.62 (0.52-0.71) in the training and internal validation sets. Notably, patients with PF after prior spine SBRT had a 61% rate of PF with subsequent spine SBRT, compared to only 5% for patients without PF after prior spine SBRT. CONCLUSION Our internally-validated model identifies a high-risk group of patients more likely to develop PF after spine SBRT, for whom prophylactic steroids may confer the greatest benefit. Evaluation in a clinical trial is warranted.

BackgroundSufficiently accurate predictions of hospital readmissions are necessary for the allocation of scare clinical resources to reduce preventable readmissions. We describe the use of a data-driven approach that relies on machine learning algorithms to predict readmission at the time of discharge.MethodsWe employ random forests to clinical and administrative electronic health record data available from a cohort of 103,688 patients discharged from the acute inpatient settings of the University of Pennsylvania Health System between June 25th, 2011 and June 30th, 2013. We predict both 30-day all-cause readmissions and 7-day unplanned readmissions using only predictors available by the time of discharge. Using oversampling and undersampling of the different outcome classes of readmission and no readmission, we incorporate into our models the asymmetric costs of a false negative relative to a false positive from the perspective of a hospital. We calculate variable importance scores for included predictors. Our approach was derived and validated using split-sample internal validation.ResultsWe developed a machine learning-based model using random forests with a 5:1 relative cost ratio for 30-day all-cause readmissions that achieves a sensitivity of 65% and specificity of 71% on validation data, as well as a random forests model with a 20:1 cost ratio for 7-day unplanned readmissions that achieves a sensitivity of 62% and specificity of 66% on validation data. Prior health system utilization, clinical discharging service, and vital sign information were most predictive of readmissions.ConclusionBy modeling the complex relationships between many predictor variables and readmission data for a large health system, we demonstrate successful predictive models that can be used upon discharge to flag patients at high risk of readmission.

Background and purpose We performed a systematic review to evaluate available risk models to predict late seizure onset among stroke survivors. Methods We searched major databases (PubMed, SCOPUS, and Cochrane Library) from inception to October 2020 for articles on the development and/or validation of risk models to predict late seizures after a stroke. The impact of models to predict late-onset seizures was also assessed. We included seven articles in the final analysis. For each of these studies, we evaluated the study design and scope of predictors analyzed to derive each model. We assessed the performance of the models during internal and external validation in terms of discrimination and calibration. Results Three studies focused on ischemic stroke alone, with c-statistic values ranging from 0.73 to 0.77. The SeLECT model from Switzerland was externally validated in Italian, German, and Austrian cohorts where c-statistics ranged from 0.69 to 0.81. This model along with the PSEiCARe model, were internally validated and calibration performance was provided for both models. The CAVS and CAVE models reported on the risk of late-onset seizures in patients with hemorrhagic stroke. The CAVS model derivation cohort was racially diverse. The CAVS model’s c-statistic was 0.76, while the CAVE model had a c-statistic of 0.81. Calibration and internal validation were not performed for either study. The CAVS model, created from a Finnish population, was externally validated in American and French cohorts, with c-statistics of 0.73 and 0.69, respectively. Finally, the two studies focusing on both types of stroke came from the PoSERS and INPOSE models. Neither model provided c-statistics, calibration metrics, internal or external validation information. We found no evidence of the presence of impact studies to assess the effect of adopting late-onset seizure risk models after stroke on clinical outcomes. Conclusion The SeLECT model was the only model developed in line with proposed guidelines for appropriate model development. The model, which was externally validated in a very similar and homogeneous population, may need to be tested in a more racially/ethnic diverse and younger population; testing the SeLECT model, accounting for overall brain health is likely to improve the identification of high-risk patients for late post stroke seizures.

Humeral torsion (HT) has been linked to pitching arm injury risk after controlling for shoulder range of motion. Currently measuring HT uses expensive equipment, which inhibits clinical assessment. Developing an HT predictive model can aid clinical baseball arm injury risk examination. Therefore, the purpose of this study was to develop and internally validate an HT prediction model using standard clinical tests and measures in professional baseball pitchers.An 11-year (2009-2019) prospective professional baseball cohort was used for this study. Participants were included if they were able to participate in all practices and competitions and were under a Minor League Baseball contract. Preseason shoulder range of motion (external rotation [ER], internal rotation [IR], horizontal adduction [HA]) and HT were collected each season. Player age, arm dominance, arm injury history, and continent of origin were also collected. Examiners were blinded to arm dominance. An a priori power analysis determined that 244 players were needed for accurate prediction models. Missing data was low (3%); thus, a complete case analysis was performed. Model development followed the transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) recommendations. Regression models with restricted cubic splines were performed. Following primary model development, bootstrapping with 2000 iterations were performed to reduce overfitting and assess optimism shrinkage. Prediction model performance was assessed through root mean square error (RMSE), RA total of 407 professional pitchers (age: 23.2 [standard deviation 2.4] years, left-handed: 17%; arm history prevalence: 21%) participated. Predictors with the highest influence within the model include IR (0.4, 95% CI 0.3, 0.5; P.001), ER (-0.3, 95% CI -0.4, -0.2; P.001), HA (0.3, 95% CI 0.2, 0.4; P.001), and arm dominance (right-handed: -1.9, 95% CI -3.6, -0.1; P = .034). Final model RMSE was 12, REvery 3° of IR explained 1° of HT. Every 3° of ER explained 1° less of HT, and every 7° of HA explained 1° of HT. Right-handers had 2° less HT. Models demonstrated good predictive performance. This predictive model can be used by clinicians to infer HT using standard clinical test and measures. These data can be used to enhance professional baseball arm injury examination.

Objectives To develop nomograms predicting overall survival (OS), freedom from locoregional recurrence (FFLR), and freedom from distant metastasis (FFDM) for patients receiving chemoradiation for laryngeal squamous cell carcinoma (LSCC). Material and methods Clinical and treatment data for patients with LSCC enrolled on NRG Oncology/RTOG 0129 and 0522 were extracted from the RTOG database. The dataset was partitioned into 70% training and 30% independent validation datasets. Significant predictors of OS, FFLR, and FFDM were obtained using univariate analysis on the training dataset. Nomograms were built using multivariate analysis with four a priori variables (age, gender, T-stage, and N-stage) and significant predictors from the univariate analyses. These nomograms were internally and externally validated using c-statistics (c) on the training and validation datasets, respectively. Results The OS nomogram included age, gender, T stage, N stage, and number of cisplatin cycles. The FFLR nomogram included age, gender, T-stage, N-stage, and time-equivalent biologically effective dose. The FFDM nomogram included age, gender, N-stage, and number of cisplatin cycles. Internal validation of the OS nomogram, FFLR nomogram, and FFDM nomogram yielded c = 0.66, c = 0.66 and c = 0.73, respectively. External validation of these nomograms yielded c = 0.59, c = 0.70, and c = 0.73, respectively. Using nomogram score cutoffs, three risk groups were separated for each outcome. Conclusions We have developed and validated easy-to-use nomograms for LSCC outcomes using prospective cooperative group trial data.

Carbonized traditional Chinese medicine (TCM) is a kind of distinctive traditional drug which has been widely used in various bleeding syndromes for over two thousand years, and most of them are still in clinical use. Although they share similar processing method: stir-frying, there are no specific quality standards and few quality control researches carried out on carbonized TCM up until now. Carbonized Typhae Pollen (CTP) is a typical carbonized TCM with efficacy of eliminating blood stasis and stanching bleeding. In this study, a novel process quality control model coupled with near infrared spectroscopy was established, called Gradient-based Discriminant Analysis method (GDA). Compared with conventional modeling methods (Convolutional Neural Network, Linear Discriminant Analysis, Standard Normal Variate-LDA), GDA model applied in fiber optic probe acquisition mode exhibited highest test accuracy (0.961), satisfactory correct identification (internal validation, 100%; external validation, 97.1%) and excellent model stability. This method provided a perfect guideline for process quality control of Carbonized TCM as well as ensured their clinical efficacy.

Background and aimsHighly accurate noninvasive methods for predicting gastroesophageal varices needing treatment (VNT) are desired. Radiomics is a newly emerging technology of image analysis. This study aims to develop and validate a novel noninvasive method based on radiomics for predicting VNT in cirrhosis.MethodsIn this retrospective-prospective study, a total of 245 cirrhotic patients were divided as the training set, internal validation set and external validation set. Radiomics features were extracted from portal-phase computed tomography (CT) images of each patient. A radiomics signature (Rad-score) was constructed with the least absolute shrinkage and selection operator algorithm and 10-folds cross-validation in the training set. Combined with independent risk factors, a radiomics nomogram was built with a multivariate logistic regression model. ResultsThe rad-score, consisting of 14 features from the gastroesophageal region and 5 from the splenic hilum region, was effective for VNT classification. The diagnostic performance was further improved by combining the rad-score with platelet counts, achieving an AUC of 0.987(95% CI, 0.969-1.00), 0.973(95% CI, 0.939-1.00) and 0.947(95% CI, 0.876-1.00) in the training set, internal validation set and external validation set respectively. In efficacy and safety assessment, the radiomics nomogram could spare more than 40% of endoscopic examinations with a low risk of missing VNT (

Purpose Machine learning (ML) is a strong candidate for making accurate predictions, as we can use large amount of data with powerful computational algorithms. We developed a ML based model to predict survival of patients with colorectal cancer (CRC) using data from two independent datasets.Materials and Methods A total of 364,316 and 1,572 CRC patients were included from the Surveillance, Epidemiology, and End Results (SEER) and a Korean dataset, respectively. As SEER combines data from 18 cancer registries, internal validation was done using 18-Fold-Cross-Validation then external validation was performed by testing the trained model on the Korean dataset. Performance was evaluated using area under the receiver operating characteristic curve (AUROC), sensitivity and positive predictive values.Results Clinicopathological characteristics were significantly different between the two datasets and the SEER showed a significant lower 5-year survival rate compared to the Korean dataset (60.1% vs. 75.3%, p < 0.001). The ML-based model using the Light gradient boosting algorithm achieved a better performance in predicting 5-year-survival compared to American Joint Committee on Cancer stage (AUROC, 0.804 vs. 0.736; p < 0.001). The most important features which influenced model performance were age, number of examined lymph nodes, and tumor size. Sensitivity and positive predictive values of predicting 5-year-survival for classes including dead or alive were reported as 68.14%, 77.51% and 49.88%, 88.1% respectively in the validation set. Survival probability can be checked using the web-based survival predictor (http://colorectalcancer.pythonanywhere.com).Conclusion ML-based model achieved a much better performance compared to staging in individualized estimation of survival of patients with CRC.

Introduction This study aims to construct a real-time deep convolutional neural networks (DCNNs) system to diagnose early esophageal squamous cell carcinoma (ESCC) with white light imaging endoscopy. Methods A total of 4,002 images from 1,078 patients were used to train and cross-validate the DCNN model for diagnosing early ESCC. The performance of the model was further tested with independent internal and external validation data sets containing 1,033 images from 243 patients. The performance of the model was then compared with endoscopists. The accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and Cohen kappa coefficient were measured to assess performance. Results The DCNN model had excellent performance in diagnosing early ESCC with a sensitivity of 0.979, a specificity of 0.886, a positive predictive value of 0.777, a negative predictive value of 0.991, and an area under curve of 0.954 in the internal validation data set. The model also depicted a tremendously generalized performance in 2 external data sets and exhibited superior performance compared with endoscopists. The performance of the endoscopists was markedly elevated after referring to the predictions of the DCNN model. An open-accessed website of the DCNN system was established to facilitate associated research. Discussion A real-time DCNN system, which was constructed to diagnose early ESCC, showed good performance in validation data sets. However, more prospective validation is needed to understand its true clinical significance in the real world.

Background: This study aimed to establish and validate a novel clinical model to differentiate between benign and malignant solitary pulmonary nodules (SPNs). Methods: Records from 406 patients with SPNs in Sun Yat-sen University Cancer Center were retrospectively reviewed. The data was randomly divided into training cohort and internal validation cohort. Other 190 SPNs patients in Henan Tumor Hospital were used for external validation. The novel prediction model was established using LASSO logistic regression analysis by integrating clinical features, radiologic characteristics and laboratory test data, the calibration of model was analyzed using the Hosmer-Lemeshow test (HL test). Subsequently, the model was compared with PKUPH and Mayo models using receiver-operating characteristics curve (ROC), decision curve analysis (DCA), net reclassification improvement index (NRI), and integrated discrimination improvement index (IDI) with the same data. Findings: A total of 15 variables were screened out and then aggregated to generate new prediction model. The model showed good calibration with the HL test (P = 0.221). The AUC for our model was 0.799, which was higher than other two reported models. DCA also showed our model was superior to the other two reported models. In our model, sensitivity = 70.06%, specificity = 77.08%. Compared with the PKUPH and Mayo models, the NRI of our model increased by 0.301 and 0.469 respectively, and the IDI improved 0.011 and 0.123, respectively. Furthermore, the model was significant positive correlation with PKUPH and Mayo models. Interpretation: The novel model in our study had a high clinical value in diagnose of MSPNs. Funding Statement: This work was supported by National Natural Science Foundation of China (No.81871711) and National Key Research and Development Program (2018YFC0910202). Declaration of Interests: The authors declare no potential conflicts of interest. Ethics Approval Statement: This study was approved by the Hospital Ethics Committee in Sun Yat-sen University Cancer Center and Henan Tumor Hospital. This study was conducted according to the Declaration of Helsinki.

Summary: Background: Detailed evaluation of bile duct (BD) is main focus during endoscopic ultrasound (EUS). The aim of this study was to develop a system for EUS BD scanning augmentation. Methods: The scanning was divided into 4 stations. We developed a station classification model and a BD segmentation model with 10681 images and 2529 images, respectively. 1704 images and 667 images were applied to classification and segmentation internal validation. For classification and segmentation video validation, 264 and 517 videos clips were used. For man-machine contest, an independent data set contained 120 images was applied. 799 images from other two hospitals were used for external validation. A crossover study was conducted to evaluate the system effect on reducing difficulty in ultrasound images interpretation. Findings: For classification, the model achieved an accuracy of 93.3% in image set and 90.1% in video set. For segmentation, the model had a dice of 0.77 in image set, sensitivity of 89.48% and specificity of 82.3% in video set. For external validation, the model achieved 82.6% accuracy in classification. In man-machine contest, the models achieved 88.3% accuracy in classification and 0.72 dice in BD segmentation, which is comparable to that of expert. In the crossover study, trainees’ accuracy improved from 60.8% to 76.3% (P

BackgroundTumor mutational burden (TMB) could be a measure of response to immune checkpoint inhibitors therapy for patients with colorectal cancer (CRC). MicroRNAs (miRNAs) participate in anticancer immune responses. In the present study, we determined miRNA expression patterns in patients with CRC and built a signature that predicts TMB.MethodsNext generation sequencing (NGS) on formalin-fixed paraffin-embedded samples from CRC patients was performed to measure TMB levels. We used datasets from The Cancer Genome Atlas to compare miRNA expression patterns in samples with high and low TMB from patients with CRC. We created an miRNA-based signature index using the selection operator (LASSO) and least absolute shrinkage method from the training set. We used an independent test set as internal validation. We used real-time polymerase chain reaction (RT-PCR) to validate the miRNA-based signature classifier.ResultsTwenty-seven samples from CRC patients underwent NGS to determine the TMB level. We identified four miRNA candidates in the training set for predicting TMB (N = 311). We used the test set (N = 204) for internal validation. The four-miRNA-based signature classifier was an accurate predictor of TMB, with accuracy 0.963 in the training set. In the test set, it was 0.902; and it was 0.946 in the total set. The classifier was superior to microsatellite instability (MSI) for predicting TMB in TCGA dataset. In the validation cohort, MSI status more positively correlated with TMB levels than did the classifier. Validation from RT-qPCR showed good target discrimination of the classifier for TMB prediction.ConclusionTo our knowledge, this is the first miRNA-based signature classifier validated using high quality clinical data to accurately predict TMB level in patients with CRC.

Surgical resection remains primary curative treatment for patients with hepatocellular carcinoma (HCC) while over 50% of patients experience recurrence, which calls for individualized recurrence prediction and early surveillance. This study aimed to develop a machine learning prognostic model to identify high-risk patients after surgical resection and to review importance of variables in different time intervals. The patients in this study were from two centers including Eastern Hepatobiliary Surgery Hospital (EHSH) and Mengchao Hepatobiliary Hospital (MHH). The best-performed model was determined, validated, and applied to each time interval (0–1 year, 1–2 years, 2–3 years, and 3–5 years). Importance scores were used to illustrate feature importance in different time intervals. In addition, a risk heat map was constructed which visually depicted the risk of recurrence in different years. A total of 7,919 patients from two centers were included, of which 3,359 and 230 patients experienced recurrence, metastasis or died during the follow-up time in the EHSH and MHH datasets, respectively. The XGBoost model achieved the best discrimination with a c-index of 0.713 in internal validation cohort. Kaplan-Meier curves succeed to stratify external validation cohort into different risk groups (p < 0.05 in all comparisons). Tumor characteristics contribute more to HCC relapse in 0 to 1 year while HBV infection and smoking affect patients’ outcome largely in 3 to 5 years. Based on machine learning prediction model, the peak of recurrence can be predicted for individual HCC patients. Therefore, clinicians can apply it to personalize the management of postoperative survival.

The aim of this study was to develop a predictive model of pediatric mortality in the early stages of intensive care unit (ICU) admission using machine learning. Patients less than 18 years old who were admitted to ICUs at four tertiary referral hospitals were enrolled. Three hospitals were designated as the derivation cohort for machine learning model development and internal validation, and the other hospital was designated as the validation cohort for external validation. We developed a random forest (RF) model that predicts pediatric mortality within 72 h of ICU admission, evaluated its performance, and compared it with the Pediatric Index of Mortality 3 (PIM 3). The area under the receiver operating characteristic curve (AUROC) of RF model was 0.942 (95% confidence interval [CI] = 0.912–0.972) in the derivation cohort and 0.906 (95% CI = 0.900–0.912) in the validation cohort. In contrast, the AUROC of PIM 3 was 0.892 (95% CI = 0.878–0.906) in the derivation cohort and 0.845 (95% CI = 0.817–0.873) in the validation cohort. The RF model in our study showed improved predictive performance in terms of both internal and external validation and was superior even when compared to PIM 3.

Background A shortage of donor organs amid a high demand for transplantable organs is a worldwide problem, and an increase in organ donation would be welcomed by the global healthcare system. Patients with brain death (BD) are potential organ donors, and early prediction of BD patients may facilitate the process of organ procurement. Therefore, we developed a model for early prediction of BD in patients who survived the initial phase of out-of-hospital cardiac arrest (OHCA). Methods We retrospectively analysed data of patients aged, who were aged < 80 years, experienced OHCA with return of spontaneous circulation (ROSC), and were admitted to our hospital between 2006 and 2018. We categorised the patients into either a non-BD or BD group. Demographic and laboratory data on emergency department admission were used for stepwise logistic regression. Prediction scores of BD after OHCA were based on β-coefficients of prognostic factors identified in the multivariable logistic model. Results Overall, 419 OHCA patients with ROSC were admitted to our hospital during the study period. Seventy-seven patients showed BD (18.3%). Age and OHCA aetiology were significantly different between the groups. Logistic regression analysis confirmed that age, low-flow time, pH, and aetiology were independent predictors of BD. The area under the receiver operating characteristic curve for this model was 0.831 (95% CI, 0.786–0.876). Conclusion We developed and internally validated a new prediction model for BD after OHCA, which could aid in early identification of potential organ donors for early donor organ procurement.

Despite the current hype in almost every area of the society and medicine, artificial intelligence (AI) is not very new. Machine learning (ML), one of the first AI methods, was described in the 50s, followed recently by an extraordinary technical development of computers and processing power leading to the current achievements of deep learning (DL). DL is the current state-of-art technique in ML, particularly well adapted for image analysis. DL techniques have allowed new algorithms to make predictions on various outcomes and diagnosis, based on « learning » (training) on large datasets, using either a supervised (labelled data) or un-supervised (unlabelled data) approach. As in other areas of DL, the input data is used for training purposes and needs to be selected according to the highest clinical diagnostic standards, because it represents the « ground truth » or the « reference standard ». All the results provided by a DL algorithm will be compared to this ground truth. After the training phase, the performance of the algorithm is tested, first via internal validation (cross-validation) and, more importantly, via external testing (in totally novel datasets). The datasets used for training, internal validation and external testing need to be distinct and should not intersect at any time. The performance of an algorithm is expressed as diagnostic accuracy, sensitivity, specificity and area under the receiving operating curve (AUC), compared to the reference standard. The current, extraordinary boom of AI requires a high need of rigorous, controlled, prospective evaluations to demonstrate the impact of AI systems in health outcomes. In response to this need, AI consensus groups have recently elaborated international guidelines regarding AI interventions, including instructions and skills required for use of AI systems, including considerations for the handling of input and output data, the human–AI interaction and analysis of error cases [1].

Background: Knowledge of predictors for shoulder pain in swimmers can assist professionals working with the athlete in developing optimal prevention strategies. However, study methodology and limited available data have constrained a comprehensive understanding of which factors cause shoulder pain. Purpose: To investigate risk factors and develop and internally validate a multivariable prognostic model for the prediction of shoulder pain in swimmers. Study Design: Cohort study; Level of evidence, 2. Methods: A total of 201 pain-free club- to international-level competitive swimmers were followed for 2 consecutive seasons. The cohort consisted of 96 male (mean ± SD age, 13.9 ± 2.2 years) and 105 female (13.9 ± 2.2 years) swimmers. Demographic, sport-specific, and musculoskeletal characteristics were assessed every 6 months. Swim-training exposure was observed prospectively. Shoulder pain interfering with training was the primary outcome. Multiple imputation was used to cope with missing data. The final model was estimated using multivariable logistic regression. We applied bootstrapping to internally validate the model and correct for overoptimism. Results: A total of 42 new cases of shoulder pain were recorded during the study. Average duration of follow-up was 1.1 years. Predictors included in the final model were acute:chronic workload ratio (odds ratio [OR], 4.31; 95% CI, 1.00-18.54), competitive level (OR, 0.19; 95% CI, 0.06-0.63), shoulder flexion range of motion, posterior shoulder muscle endurance (OR, 0.96; 95% CI, 0.92-0.99), and hand entry position error (OR, 0.37; 95% CI, 0.16-0.91). After internal validation, this model maintained good calibration and discriminative power (area under the receiver operating characteristic curve, 0.71; 95% CI, 0.60-0.94). Conclusion: Our model consists of parameters that are readily measurable in a swimming setting, allowing the identification of swimmers at risk for shoulder pain. Multivariable logistic regression showed the strongest predictors for shoulder pain were regional competitive swimming level, acute:chronic workload ratio, posterior shoulder muscle endurance, and hand entry error.

Background: The evaluation of tumor differentiation is an urgent clinical issue that would facilitate the establishment of individualized therapeutic strategies. Our aims were to develop a deep learning radiomics model based on computed tomography (CT) data for preoperative evaluation of hepatocellular carcinoma (HCC) differentiation (low vs high grade), and to preliminarily explore the biological basis of the radiomics model. Methods: A total of 1234 HCC patients with contrast-enhanced CT images were recruited from two institutions. Radiomics features were extracted from preoperative venous-phase CT data and selected in terms of reproducibility, relevance, and redundancy. The random forest (RF) algorithm was applied to establish a radiomics signature. Meanwhile, the deep learning model was constructed based on a VGG network. Additionally, clinical characteristics of the subjects were used to construct a clinical model. Finally, the fused model integrated a tripartite prediction based on a logistic regression algorithm. A correlation analysis was performed to explore the association between radiomics features and biological variables based on multiomics levels. Findings: The radiomics signature established with the RF algorithm comprised 25 radiomics features. The AUCs in the training, internal validation, and independent test cohorts were 0·82, 0·76, and 0·75, respectively, for the radiomics signature; 0·85, 0·81, and 0·75, respectively, for the deep learning model; and 0·89, 0·83, and 0·80, respectively, for the fused model. Multiomics analysis showed that the selected radiomics features contained abundant biological information that was related to tumor differentiation. Interpretation: Our deep learning radiomics model can serve as a noninvasive tool of preoperative HCC differentiation evaluation to guide clinical decision-making and prognostic stratification. Funding: National Key Research and Development Program of China, Natural Science Foundation of China, Zhejiang Provincial Education Department. Declaration of Interest: The authors declare that they have no competing interests. Ethical Approval: This study was approved by the institutional review boards of the First Affiliated Hospital, College of Medicine, Zhejiang University (Institution 1) and Ningbo Medical Center Lihuili Hospital (Institution 2), which waived the requirement for patients’ informed consent.

With the recent advances in next-generation sequencing (NGS), mitochondrial whole-genome sequencing has begun to be applied to the field of the forensic biology as an alternative to the traditional Sanger-type sequencing (STS). However, experimental workflows, commercial solutions, and output data analysis must be strictly validated before being implemented into the forensic laboratory. In this study, we performed an internal validation for an NGS-based typing of the entire mitochondrial genome using the Precision ID mtDNA Whole Genome Panel (Thermo Fisher Scientific) on the Ion S5 sequencer (Thermo Fisher Scientific). Concordance, repeatability, reproducibility, sensitivity, and heteroplasmy detection analyses were assessed using the 2800 M and 9947A standard control DNA as well as typical casework specimens, and results were compared with conventional Sanger sequencing and another NGS sequencer in a different laboratory. We discuss the strengths and limitations of this approach, highlighting some issues regarding noise thresholds and heteroplasmy detection, and suggesting solutions to mitigate these effects and improve overall data interpretation. Results confirmed that the Precision ID Whole mtDNA Genome Panel is highly reproducible and sensitive, yielding useful full mitochondrial DNA sequences also from challenging DNA specimens, thus providing further support for its use in forensic practice. Supplementary Information The online version contains supplementary material available at 10.1007/s00414-021-02686-w.

Background Prognostication of neurological outcome in patients who remain comatose after cardiac arrest resuscitation is complex. Clinical variables, as well as biomarkers of brain injury, cardiac injury, and systemic inflammation, all yield some prognostic value. We hypothesised that cumulative information obtained during the first three days of intensive care could produce a reliable model for predicting neurological outcome following out-of-hospital cardiac arrest (OHCA) using artificial neural network (ANN) with and without biomarkers. Methods We performed a post hoc analysis of 932 patients from the Target Temperature Management trial. We focused on comatose patients at 24, 48, and 72 h post-cardiac arrest and excluded patients who were awake or deceased at these time points. 80% of the patients were allocated for model development (training set) and 20% for internal validation (test set). To investigate the prognostic potential of different levels of biomarkers (clinically available and research-grade), patients’ background information, and intensive care observation and treatment, we created three models for each time point: (1) clinical variables, (2) adding clinically accessible biomarkers, e.g., neuron-specific enolase (NSE) and (3) adding research-grade biomarkers, e.g., neurofilament light (NFL). Patient outcome was the dichotomised Cerebral Performance Category (CPC) at six months; a good outcome was defined as CPC 1–2 whilst a poor outcome was defined as CPC 3–5. The area under the receiver operating characteristic curve (AUROC) was calculated for all test sets. Results AUROC remained below 90% when using only clinical variables throughout the first three days in the ICU. Adding clinically accessible biomarkers such as NSE, AUROC increased from 82 to 94% (p Conclusions In this exploratory study, ANNs provided good to excellent prognostic accuracy in predicting neurological outcome in comatose patients post OHCA. The models which included NSE after 72 h and NFL on all days showed promising prognostic performance.

Powdery mildew (Podosphaera aphanis) is a major disease in day-neutral strawberry. Up to 30% yield losses have been observed in Eastern Canada. Currently, management of powdery mildew is mostly based on fungicide applications without consideration of risk. The objective of this study is to use P. aphanis inoculum, host ontogenic resistance, and weather predictors to forecast the risk of strawberry powdery mildew using CART models (classification trees). The data used to build the trees were collected in 2006, 2007, and 2008 at one experimental farm and six commercial farms located in two main strawberry-production areas, while external validation data were collected at the same experimental farm in 2015, 2016, and 2018. Data on proportion of leaf area diseased (PLAD) were grouped into four severity classes (1: PLAD = 0, 2: PLAD &gt, 0 and &lt, 5%, 3: &gt, 5% and &lt, 15%, and 4: PLAD &gt, 15%) for a total of 681 and 136 cases for training and external validation, respectively. From the initial 92 weather variables, 21 were selected following clustering. The tree with the best balance between the number of predictors and highest accuracy was built with: airborne inoculum concentration and number of susceptible leaves on the day of sampling, and mean relative humidity, mean daily number of hours at temperature between 18 and 30 &deg, C, and mean daily number of hours at saturation vapor pressure between 10 and 25 mmHg during the previous 6 days. For training, internal validation, and external validation datasets, the sensitivity, specificity, and accuracy ranged from 0.70 to 0.90, 0.87 to 0.98, and 0.82 to 0.97, respectively. The classification rules to estimate strawberry powdery mildew risk can be easily implemented into disease decision support systems and used to treat only when necessary and thus avoid preventable yield losses and unnecessary treatments.

BackgroundHypercortisolism accounts for relevant morbidity and mortality and is often a diagnostic challenge for clinicians. A prompt diagnosis is necessary to treat Cushing’s syndrome as early as possible.ObjectiveThe aim of this study was to develop and validate a clinical model for the estimation of pre-test probability of hypercortisolism in an at-risk population.DesignWe conducted a retrospective multicenter case-control study, involving five Italian referral centers for Endocrinology (Turin, Messina, Naples, Padua and Rome). One hundred and fifty patients affected by Cushing’s syndrome and 300 patients in which hypercortisolism was excluded were enrolled. All patients were evaluated, according to current guidelines, for the suspicion of hypercortisolism.ResultsThe Cushing score was built by multivariable logistic regression, considering all main features associated with a clinical suspicion of hypercortisolism as possible predictors. A stepwise backward selection algorithm was used (final model AUC=0.873), then an internal validation was performed through ten-fold cross-validation. Final estimation of the model performance showed an average AUC=0.841, thus reassuring about a small overfitting effect. The retrieved score was structured on a 17.5-point scale: low-risk class (score value: ≤5.5, probability of disease=0.8%); intermediate-low-risk class (score value: 6-8.5, probability of disease=2.7%); intermediate-high-risk class (score value: 9-11.5, probability of disease=18.5%) and finally, high-risk class (score value: ≥12, probability of disease=72.5%).ConclusionsWe developed and internally validated a simple tool to determine pre-test probability of hypercortisolism, the Cushing score, that showed a remarkable predictive power for the discrimination between subjects with and without a final diagnosis of Cushing’s syndrome.

Background: Pre-existing indices predicting the prognosis of COPD do not contain radiologic information and are impractical because they involve complex history-taking or exercise tests. We aimed to develop and validate a deep learning-based survival predictor for COPD utilizing chest radiographs (DLSPCXR) and integrate it with other clinical factors (DLSPinteg). Methods: Patient with COPD who underwent post-bronchodilator spirometry and chest radiograph (CXR) in 2011-2015 were retrospectively collected and split into training (n=3,475), tuning (n=435), and internal validation (n=315) datasets. A deep-learning algorithm predicting survival from CXR was trained (DLSPCXR), and was integrated with clinical factors, including age, body mass index, and forced expiratory volume in 1 second (FEV1) (DLSPinteg). For external validation, three cohorts were collected from different institutions (n=337-416). The discrimination performance of DLSP was evaluated using time-dependent area-under-the receiver-operating characteristic curves (TD-AUROCs) for 5-year survival. Using one external validation cohort, DLSP integ was compared with four COPD-specific clinical indices: BODE, ADO, CAT, and SGRQ. Goodness-of-fit was assessed using the Hosmer-Lemeshow test. Findings: DLSP CXR showed TD-AUROC of 0.670-0.761 for 5-year survival, significantly higher than that of FEV1 in two of three external validation cohorts. DLSPCXR demonstrated good calibration (p>0.05) and worked as an independent survival predictor from FEV1 (p

No studies have discussed machine learning algorithms to predict recurrence within 10 years after breast cancer surgery. This study purposed to compare the accuracy of forecasting models to predict recurrence within 10 years after breast cancer surgery and to identify significant predictors of recurrence. Registry data for breast cancer surgery patients were allocated to a training dataset (n = 798) for model development, a testing dataset (n = 171) for internal validation, and a validating dataset (n = 171) for external validation. Global sensitivity analysis was then performed to evaluate the significance of the selected predictors. Demographic characteristics, clinical characteristics, quality of care, and preoperative quality of life were significantly associated with recurrence within 10 years after breast cancer surgery (p &lt, 0.05). Artificial neural networks had the highest prediction performance indices. Additionally, the surgeon volume was the best predictor of recurrence within 10 years after breast cancer surgery, followed by hospital volume and tumor stage. Accurate recurrence within 10 years prediction by machine learning algorithms may improve precision in managing patients after breast cancer surgery and improve understanding of risk factors for recurrence within 10 years after breast cancer surgery.

Background and objective CGM systems are still subject to a time-delay, which especially during rapid changes causes clinically significant difference between the CGM and the actual BG level. This study had the aim of exploring the potential of developing and validating a model for prediction of future CGM measurements in order to overcome the time-delay. Methods An artificial neural network regression (NN) approach were used to predict CGM values with a lead-time of 15 min. The NN were trained and internally validated on 23 million minutes of CGM and externally validated on 2 million minutes of CGM. The validation included data from 278 type 1 diabetes patients using three different CGM sensors. The NN performance were compared with three alternative methods, linear extrapolation, spline extrapolation and last observation carried forward. Results The internal validation yielded a RMSE of 9.1 mg/dL, a MARD of 4.2 % and 99.9 % of predictions were in the A + B zone of the consensus error grid. The external validation yielded a RMSE of 5.9–11.3 mg/dL, a MARD of 3.2–5.4 % and 99.9–100 % of predictions were in the A + B zone of the consensus error grid. The NN performed better on all parameters compared to the two alternative methods. Conclusions We proposed and validated a NN glucose prediction model that is potential simple to use and implement. The model only needs input from a CGM system in order to facilitate glucose prediction with a lead time of 15 min. The approach yielded good results for both internal and external validation.

BackgroundThe status of health of an individual and - more broadly - of a community or population is defined by the status of their determinants of health. A “systemic” approach to define the health determinants is necessary in order to explore the complex relations existing among them. This study is aimed at identifying a ‘composite systemic’ index of health to measure the impact of socioeconomic factors on public health at local level and to analyze possible spatial autocorrelations between neighboring regions.ResultsA Composite Index of Health (CIH) was constructed on the basis of known indicators of socio-economic well being by using the COMIC (COMposite Indices Creator) Software and was validated on the Italian population and a nationwide comparison has been performed. Analysis of the determinants showed a significant direct correlation between health, environment, work and wealth and inverse correlation between health and social distress. The analysis of data from Italian provinces confirmed the South-North gradient of well-being.ConclusionThe CIH is a reliable and robust index to evaluate the health of a local population. Although it was validated on Italian data, the index can be easily adapted to any Country.