جستجوی مقالات مرتبط با کلیدواژه "prediction" در نشریات گروه "پزشکی"

To develop a deep learning model to predict the risk of cardiovascular diseases (CVDs) events based on features found in fundus images.

We developed a predicting model for cardiovascular diseases based on retinal fundus images using the deep learning method. We trained our model using 2,091 retinal fundus images obtained from 211 patients. Our dataset included demographic information of each person, conventional CVD risk factors, CVD risk estimated number (calculated using the Framingham method), strokes and heart attack incidents during 5 years (patients who were referred to the ICU or CCU), and retinal fundus images for each person. We used receiver operating characteristic (ROC) analysis to assess the accuracy of our classification model.

Our proposed algorithm was able to identify high-risk individuals from no-risk individuals with 83% accuracy and a high confidence level (AUC = 0.91, P value < 0.0001). The results also showed that our model could predict cardiovascular events such as stroke with a probability of 72% (AUC = 0.83, P value< 0.0001). In comparing our model’s ability to predict CVD risk with the Framingham risk score, the Framingham model’s accuracy was 65 % in our dataset (with a best AUC of 0.78).

Our deep learning prediction model developed based on retinal fundus image findings to predict the risk of CVD, showed a relatively high accuracy. Its accuracy was higher than traditional prediction models like the Framingham model and comparable to other models based on fundus images for predicting CVD.

Transfusion of packed red blood cells (PRBCs) following severe bleeding from multiple trauma can reduce mortality. The present study aimed to compare the accuracy of eight different scoring systems for predicting the need for blood transfusion in such patients. The present diagnostic accuracy study was conducted at the emergency department of Shohadaye Tajrish Hospital in Tehran, From March to September 2023. Medical records of multiple trauma patients admitted to the emergency department were reviewed. The predictive performances of eight scoring systems including Glasgow coma scale (GCS), revised trauma score (RTS), trauma associated severe hemorrhage (TASH), Prince of Wales hospital score (PWH), emergency transfusion score (ETS), base deficit, assessment of blood consumption (ABC), and the Shock index in predicting the need for PRBC transfusion were assessed. The area under the ROC curve of TASH in predicting PRBC transfusion was calculated 0.959, significantly higher than the area under the ROC curves for PWH, GCS, Shock index, Base deficit, ETS, RTS, and ABC (0.902, 0.899, 0.882, 0.857, 0.846, 0.824, and 0.810 respectively; p < 0.001). Sensitivity and specificity of TASH at the optimal cut-off were 98.72% and 51.56% respectively. A new score, the MTTP (Multiple Trauma Transfusion Predictor), developed by evaluating the association of clinical and laboratory variables with PRBC transfusion in the ED, showed an AUC of 0.964, not significantly higher than the AUC of TASH (p=0.804). The sensitivity and specificity of MTTP at the optimal cut-off were 93.59% and 91.84%, respectively. Among the evaluated scores, TASH was the most accurate for predicting PRBC transfusion in multiple trauma patients in the ED. Furthermore, among the pre-hospital scores, the Shock index was identified as the most accurate predictor for PRBC transfusion. This score is recommended for use in the ED due to its simplicity, rapid calculation and high prediction accuracy. The MTTP, the newly developed scoring system in this study, outperformed all the other scores.

Breast cancer represents a major public health issue worldwide, highlighting the critical role of early detection in facilitating effective treatment. Fine needle aspiration (FNA) serves as a minimally invasive method for obtaining cellular material from breast masses for subsequent analysis. Nonetheless, pathologists' assessment of FNA samples may be characterized by subjectivity and protracted evaluation times, leading to variability in diagnostic results. Integrating machine learning algorithms, including classification tree models, can potentially improve the consistency and precision of breast tumor classification. Using computational capabilities and sophisticated machine learning methodologies, these models can proficiently categorize digitized images of FNA samples as malignant or benign.

We used classification tree algorithms such as CART, Ctree, Evtree, QUEST, CRUISE, and GUIDE to distinguish between malignant and benign tumors in the Wisconsin Breast Cancer Dataset (WBCD). The models' performance was evaluated using accuracy metrics, such as sensitivity, specificity, false positive and negative rates, positive and negative predictive values, Youden's Index, accuracy, positive and negative likelihood ratios, diagnostic odds ratios, and AUC (area under the ROC curve).

The results showed that the CRUISE algorithm showed excellent diagnostic performance in distinguishing between malignant and benign tumors.

The results emphasize the critical role of integrating machine learning models into clinical practice to assist pathologists, improve diagnostic outcomes, and reduce subjectivity in cancer classification.

The COVID-19 pandemic has had a significant impact on global health, resulting in more than 6 million reported deaths worldwide as of April 2023. This study aimed to investigate the potential of C-reactive protein (CRP), procalcitonin (PCT), and D-dimer as biomarkers for assessing disease severity in COVID-19 patients in Kinshasa, Democratic Republic of Congo.

A retrospective examination was conducted involving 339 COVID-19 patients admitted to Kinshasa hospitals between January 2021 and March 2022. CRP, PCT, and D-dimer levels were measured in all patients and compared between those with severe and non-severe illnesses.

Our findings revealed significantly higher CRP, PCT, and D-dimer levels in severe cases compared to non-severe cases. Specifically, the median CRP level was 120.6 mg/L in severe cases, 47.3 mg/L in mild cases, and 13.5 mg/L in moderate cases. The median PCT levels were 0.26 ng/mL in severe cases, 0.08 ng/mL in mild cases, and 0.07 ng/L in moderate cases. Additionally, the median D-dimer level was 1836.9 µg/L in severe cases and 597.6 µg/L in mild cases, with a value of 481.1 µg/L in moderate cases. System learning techniques were also employed to predict disease severity based on these biomarkers, achieving an accuracy of 97%.

Our findings suggest that CRP, PCT, and D-dimer serve as valuable biomarkers for identifying severe COVID-19 cases in Kinshasa. Furthermore, the application of machine learning methods can yield accurate predictions of disease severity based on these biomarkers. These biomarkers hold the potential to assist clinicians in informed decision-making regarding patient management and contribute to improved clinical outcomes for COVID-19 patients.

In trauma patients, various severity scoring indices have been developed to predict the severity of injury and mortality. This study aimed to investigate the relationship between severity scoring indices for predicting survival in moderate and severe trauma patients. This cross-sectional study was conducted among 100 trauma patients. Information on each of the Shock Index, Glasgow Coma Scale/Age/Pressure (GAP), RGAP, New Trauma Score (NTS), Mechanism/Glasgow Coma Scale/Age/Pressure (MGAP), Modified Early Warning Score (MEWS), and Trauma and Injury Severity Score (TRISS) indices was separately completed for moderate and severe trauma patients. Statistical analyses were performed using SPSS version 21. In this study, no significant differences were observed between surviving and deceased patients in terms of the Shock Index, GAP, RGAP, NTS, MGAP, MEWS, and TRISS indices. However, in all these indices, significant differences were observed between multi-trauma patients with and without morbidity. According to the ROC curve, a value less than 92.5% for the SPO2 variable was the best cutoff point for predicting the probability of death in multi-trauma patients. Also, ROC curves showed that a value higher than 95.5% for SPO2, a GCS score less than 7.5, an RR value less than 19.5, a GAP score less than 14.5, an RGAP score less than 13.5, an MGAP score less than 18, a TRISS score less than 54.2, an MEWS score higher than 3.5, and an NTS score less than 14.5 were the best ways to tell if a patient with multiple injuries was likely to be hospitalized. The GAP, RGAP, MGAP, NTS, and TRISS scoring systems performed well in predicting morbidity in multi-trauma patients. However, according to the ROC curve, the GAP and RGAP indices performed slightly better than other indices. The findings of this study can be useful in better assessing survival rates, mortality, and timely treatment and intervention in trauma patients.

Numerous scoring systems have been developed to assess the risk associated with upper gastrointestinal bleeding (UGIB), and several studies have investigated their comparative accuracy in predicting patient outcomes. This study was undertaken to compare four well-known scoring systems, namely the pre-endoscopy Rockall score, full Rockall score, Glasgow-Blatchford Bleeding score (GBS), and AIMS65, with the aim of predicting five distinct outcomes in cases of non-variceal UGIB. 

This prospective observational study was conducted focusing on adult patients with UGIB presenting to the emergency department (ED). The primary endpoints of this study included in-hospital mortality, the need for re-endoscopy, the requirements for packed red blood cell (PRBC) transfusion, massive transfusion, and one-month rebleeding. 

A total number of 320 patients were enrolled, with 44 (13·75%) in-hospital deaths. Based on the area under the curves (AUC), while certain scores outperformed others in specific outcome prediction, the AIMS65 scoring system demonstrated superior predictive capability for both in-hospital mortality (0.91) and massive transfusion (0.71). Regarding PRBC transfusion requirements, both AIMS65 and GBS exhibited similar predictive capacities (AUC=0.67 and 0.68, respectively). In terms of re-endoscopy and one-month rebleeding, the GBS scoring system displayed slightly better performance compared to the other systems (AUC=0.61 and 0.63, respectively). In the composite outcome, all scores had significant associations, and among them, the AIMS-65 score had the highest AUC (0.76). 

The AIMS65 scoring system was the most reliable tool for predicting in-hospital mortality and, to a lesser extent, massive transfusion requirements, while GBS and AIMS65 could be moderately and cautiously relied on for preparations regarding the need for PRBC transfusion.

The recent novel coronavirus disease 2019 (COVID-19) pandemic has underlined the importance of risk score models in public health emergencies. This study aimed to develop a risk prediction score to identify high-risk hospitalized patients for disease progression on admission.

This prospective cohort study included 171 COVID-19 patients, identified through the reverse transcription polymerase chain reaction test, admitted to Bohlool Hospital in Gonabad City, Iran, between April 4 and June 5, 2021. The patients’ demographic, clinical, and laboratory data were collected upon admission, and clinical outcomes were monitored until the end of the study. The discovery dataset (80% of the data) was used to develop the risk score model based on clinical and laboratory features and patient characteristics to predict COVID-19 progression. An additive risk score model was developed based on the regression coefficients of the significant variables in a multiple logistic regression model. The performance of the risk score model was evaluated on the validation dataset (20% of the data) using the receiver operating characteristic (ROC) curve. Statistical analyses were performed with SPSS software, version 21.

The Mean±SD for age of participants was 59.54±20.52 years, and 48.6% were male. Most patients (82.5%) fully recovered or showed improvement, while 5.2% experienced disease progression and 12.3% died. Three variables, interleukin-6, neutrophil-to-lymphocyte ratio, and lung involvement, were found to be significant in predicting risk, with a good discriminatory ability, having an area under the ROC curve of 0.970 (95% CI, 0.935%, 1.00%) in the discovery set and 0.973 (95% CI, 0.923%, 1.00%) in the validation set.

The developed risk score model in this study can be used as a clinical diagnostic tool to identify COVID-19 patients at higher risk of disease progression and aid in informed decision-making and resource utilization in similar situations, such as respiratory disease outbreaks in the post-corona era.

Although several studies have been published about COVID-19, ischemic stroke is known yet as a complicated problem for COVID-19 patients. Scientific reports have indicated that in many cases, the incidence of stroke in patients with COVID-19 leads to death. 

The obtained mathematical equation in this study can help physicians’ decision-making about treatment and identification of influential clinical factors for early diagnosis. 

In this retrospective study, data from 128 patients between March and September 2020, including their demographic information, clinical characteristics, and laboratory parameters were collected and analyzed statistically. A logistic regression model was developed to identify the significant variables in predicting stroke incidence in patients with COVID-19. 

Clinical characteristics and laboratory parameters for 128 patients (including 76 males and 52 females; with a mean age of 57.109±15.97 years) were considered as the inputs that included ventilator dependence, comorbidities, and laboratory tests, including WBC, neutrophil, lymphocyte, platelet count, C-reactive protein, blood urea nitrogen, alanine transaminase (ALT), aspartate transaminase (AST) and lactate dehydrogenase (LDH). Receiver operating characteristic–area under the curve (ROC-AUC), accuracy, sensitivity, and specificity were considered indices to determine the model capability. The accuracy of the model classification was also addressed by 93.8%. The area under the curve was 97.5% with a 95% CI.

The findings showed that ventilator dependence, cardiac ejection fraction, and LDH are associated with the occurrence of stroke and the proposed model can predict the stroke effectively.

Heart disease is a major public health concern with millions of reported deaths annually. Data mining techniques have received attention in recent years as a tool aiding diagnosis and prediction of heart disease cases. This systematic review examines the application of data mining methods to cardiac disease diagnosis in order to identify specific types of heart-related disease that are diagnosed using data mining techniques as well as the most successful data mining methods.

This study involved a systematic review of IEEE, Science Direct, Google Scholar, Web of Science, Scopus and MEDLINE databases from 2008 until April 2023. Inclusion criteria were original papers that used data mining methods for heart disease diagnosis. Non-English papers, those without full text, studies conducted on animals, and other types of papers (conference abstracts and letters) were excluded from the study. All the retrieved references were then assessed by title and abstract according to PRISMA, after which full texts of relevant articles were analyzed. The final sample comprised of 47 articles.

Various classification methods have been utilized to diagnose heart-related disease using different mining tools, with genetic neural network data mining method having the highest accuracy among the studied techniques. Results show that predicting cardiac disease is the most commonly performed task. The demographic, bio-clinical, personal and exercise-related attributes, as well as other features used for classification were identified. The findings suggest that data mining methods hold great potential for detecting and preventing heart disease on both individual and population scales.

The study findings have implications for the prevention and treatment of cardiac disease, especially in high-risk individuals. Data mining methods can be widely applied to detect and prevent heart disease on a population scale, as well as supporting decisions for the most suitable treatment for individual patients to prevent death and reduce treatment costs.

This study aimed to predict the risk of mortality among COVID-19 patients in the central region of Iran by employing the Charlson Comorbidity Index (CCI), with adjustments made for age in the predictive model.

In this cross-sectional study, encompassing all probable, suspicious, and confirmed COVID-19 cases from the onset of the pandemic (55307 individuals), 3415 cases resulting in death were designated as the study group, while the survivors constituted the control group.

The Charlson Comorbidity Index revealed that over 11 percent of all patients had at least one underlying medical condition. Logistic regression analysis indicated a significantly elevated likelihood of mortality among patients with comorbidities. Specifically, individuals with a CCI score of 6 or higher were more than twice as likely to succumb to the virus compared to those without underlying diseases. Those with a score of 6 or more exhibited the highest odds ratio (OR 2.4; 95% CI 1.3-4.5).

The study findings underscore the heightened vulnerability of individuals to COVID-19 mortality, particularly among the elderly with pre-existing health conditions. The coexistence of age and comorbidities substantially increased the risk of death due to COVID-19 in this population. Consequently, targeted interventions and focused care strategies may be crucial for this high-risk demographic in pandemic management efforts.

In this note, we focus on statistical analysis and try to show the deleterious effects of inappropriate use of statistical analysis in medical research.Recently, Foji et al published an article entitled above and showed that the dermatology life quality index can predict the quality of life in patients with neurofibromatosis (1). However, these findings are doubtful due to the following reasons:1. The mean score of quality of life (the total score of the SF-36 questionnaire) is not clear.2. Although the correlation between SF-36 and DLQI can be informative, no correlation was found in the results section.3. The main aim of the study is to predict quality of life using the dermatology life quality index but there is no related model. The reported models are about the prediction of SF-36 dimensions.4. All the reported R-squares are very low (about 10%) indicating that the proposed models are not appropriate for the prediction aims.5. In regression modeling, statistical significance reflects no information regarding the prediction capability. Therefore, the interpretation of the results is not true. For more information, reading an article entitled “to explain or to predict” is highly suggested (2).6. To investigate the prediction capability of each variable, the amount of changes in the adjusted R-squares must be reported.7. In the data analysis section, it was claimed that the significant variables in simple regression were included in the multiple regression but it was not performed. For example, in the “role limitations due to emotional problems” dimension, all of the six variables are significant in the simple model, but just one variable was entered in the multivariate model.8. Finally, the linear regression is not appropriate for molding response variables with a limited range (the scores of SF-36 are between 0 and 100). The appropriate method for these outcomes is beta regression (3).In conclusion, the research hypothesis is rejected and the dermatology life quality index cannot predict the quality of life. Thus, the conclusion of Foji’s studies is not acceptable due to the fact that is not based on reported findings.Conflict of InterestNothing to declare.

Chronic kidney disease (CKD) poses a significant health burden worldwide, affecting approximately 10 - 15% of the global population. As one of the leading non-communicable diseases, CKD is a major cause of morbidity and mortality. Early identification of CKD is crucial for reducing its adverse effects on patient health. Prompt detection can significantly lessen the harmful consequences and enhance health outcomes for individuals with CKD.

This study aimed to evaluate and compare the effectiveness of various machine learning models in predicting the occurrence of CKD.

The study involved the collection of data from a sample of 400 patients. We applied the well-established cross-industry standard process (CRISP) methodology for data mining to analyze the data. As part of this process, we efficiently handled missing data using the mode approach and addressed outliers through the interquartile range (IQR) method. We utilized sophisticated techniques, such as CatBoost (CB), random forest (RF), and artificial neural network (ANN) models to predict outcomes. For evaluation, we used the receiver operating characteristic (ROC) curve and calculated the area under the curve (AUC).

An analysis of 400 patient records in this study identified that variables like serum creatinine, packed cell volume, specific gravity, and hemoglobin were most influential in predicting CKD. The results indicated that the CB and RF models surpassed the ANN in predicting the disease. Ten critical predictors were pinpointed for accurate disease prediction.

The ensemble models in this study not only showcased remarkable speed but also demonstrated superior accuracy. These findings suggest the potential of ensemble models as an effective tool for enhancing predictive performance in similar studies.

 The purpose of this study was to develop a robust and externally predictive in silico QSAR-neural network model for predicting plasma protein binding of drugs. This model aims to enhance drug discovery processes by reducing the need for chemical synthesis and extensive laboratory testing.

 A dataset of 277 drugs was used to develop the QSAR-neural network model. The model was constructed using a Filter method to select 55 molecular descriptors. The validation set’s external accuracy was assessed through the predictive squared correlation coefficient Q2 and the root mean squared error (RMSE).

The developed QSAR-neural network model demonstrated robustness and good applicability domain. The external accuracy of the validation set was high, with a predictive squared correlation coefficient Q2 of 0.966 and a root mean squared error (RMSE) of 0.063. Comparatively, this model outperformed previously published models in the literature.

 The study successfully developed an advanced QSAR-neural network model capable of predicting plasma protein binding in human plasma for a diverse set of 277 drugs. This model’s accuracy and robustness make it a valuable tool in drug discovery, potentially reducing the need for resource-intensive chemical synthesis and laboratory testing.

In recent years, artificial intelligence (AI) has revolutionized several aspects of human life. The availability of high-dimensionality datasets with progression in high-performance computing, and innovative deep learning architectures which are the subdomains of AI, have led to promising functions of AI in the medical contexts, particularly in oncology. Regarding the capacity of AI models in recognition and learning patterns as well as associations, these systems can be utilized in various aspects of cancer research including cancer diagnosis and treatment. To be precise, AI models are able to analyze medical images such as stained histopathology slides and radiology images and consequently pave the way for cancer diagnosis, grading, classification, tumor characterization, and prognosis prediction. Moreover, AI algorithms can assess a myriad of medical data to recognize patterns and make predictions about patient treatment outcomes, enabling more personalized treatment plans. Accordingly, AIassisted cancer treatment strategies have been shown to notably improve the quality of cancer treatment with chemotherapy, immunotherapy, and even radiotherapy while reducing the treatment toxicities.

This study aims to find candidates for testicular spermatozoa retrieval biomarkers among the seminal plasma exLncRNA pairs.

A set of exLncRNA pairs with the best potential biomarkers was selected and validated in 96 NOA samples. Weighted correlation network analysis (WGCNA) and Least Absolute Shrinkage and Selection Operator were used to identify possible biomarkers for these pairs (LASSO). These pairs' potential biomarkers were identified using receiver operating curves. Confusion matrices and sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), FP, false-negative rates (FNR), and F1 scores are calculated. Through F1 scores, we selected the best threshold value.

The relative differential expression of each pair in testicular spermatozoa retrieval (+) and testicular spermatozoa retrieval (-) men were validated. The six pairs displayed the best biomarker potential. Among them, CCDC37.DT-LOCI00505685 pair and LOC440934- LOCI01929088 (XR_001745218.1) pair showed the most significant potential and stability for detecting testicular spermatozoa retrieval in the selected and validated cohort.

CCDC37.DT-LOCI00505685 pair and LOC440934- LOCI01929088 (XR_001745218.1) pair have the potential to become new molecular biomarkers that could help to select clinical strategies for microdissection testicular sperm extraction.

 This study was designed to find the best statistical approach to scorpion sting predictions.

Study Design:

 A retrospective study.

 Multiple regression, seasonal autoregressive integrated moving average (SARIMA), neural network autoregressive (NNAR), and hybrid SARIMA-NNAR models were developed to predict monthly scorpion sting cases in El Oued province. The root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) were used to quantitatively compare different models.

 In general, 96909 scorpion stings were recorded in El Oued province from 2005-2020. The incidence rate experienced a gradual decrease until 2012 and since then slight fluctuations have been noted. Scorpion stings occurred throughout the year with peaks in September followed by July and August and troughs in December and January. Sting cases were not evenly distributed across demographic groups; the most affected age group was 15-49 years, and males were more likely to be stung. Of the reported deaths, more than half were in children 15 and younger. Scorpion’s activity was conditioned by climate factors, and temperature had the highest effect. The SARIMA(2,0,2)(1,1,1)12, NNAR(1,1,2)12, and SARIMA(2,0,2)(1,1,1)12-NNAR(1,1,2)12 were selected as the best-fitting models. The RMSE, MAE, and MAPE of the SARIMA and SARIMA-NNAR models were lower than those of the NNAR model in fitting and forecasting; however, the NNAR model could produce better predictive accuracy.

 The NNAR model is preferred for short-term monthly scorpion sting predictions. An in-depth understanding of the epidemiologic triad of scorpionism and the development of predictive models ought to establish enlightened, informed, better-targeted, and more effective policies.