BIO-CXRNET: a robust multimodal stacking machine learning technique for mortality risk prediction of COVID-19 patients using chest X-ray images and clinical data

Neural Computing and Applications, Journal Year: 2023, Volume and Issue: 35(24), P. 17461 - 17483

Published: May 4, 2023

Nowadays, quick, and accurate diagnosis of COVID-19 is a pressing need. This study presents multimodal system to meet this The presented employs machine learning module that learns the required knowledge from datasets collected 930 patients hospitalized in Italy during first wave (March-June 2020). dataset consists twenty-five biomarkers electronic health record Chest X-ray (CXR) images. It found can diagnose low- or high-risk with an accuracy, sensitivity, F1-score 89.03%, 90.44%, respectively. exhibits 6% higher accuracy than systems employ either CXR images biomarker data. In addition, calculate mortality risk using multivariate logistic regression-based nomogram scoring technique. Interested physicians use predict early risks web-link: Covid-severity-grading-AI. case, physician needs input following information: image file, Lactate Dehydrogenase (LDH), Oxygen Saturation (O2%), White Blood Cells Count, C-reactive protein, Age. way, contributes management by predicting risk.The online version contains supplementary material available at 10.1007/s00521-023-08606-w.

Language: Английский

---

FECNet: a Neural Network and a Mobile App for COVID-19 Recognition

Mobile Networks and Applications, Journal Year: 2023, Volume and Issue: 28(5), P. 1877 - 1890

Published: July 17, 2023

Abstract COVID-19 has caused over 6.35 million deaths and 555 confirmed cases till 11/July/2022. It a serious impact on individual health, social economic activities, other aspects. Based the gray-level co-occurrence matrix (GLCM), four-direction varying-distance GLCM (FDVD-GLCM) is presented. Afterward, five-property feature set (FPFS) extracts features from FDVD-GLCM. An extreme learning machine (ELM) used as classifier to recognize COVID-19. Our model finally dubbed FECNet. A multiple-way data augmentation method utilized boost training sets. Ten runs of tenfold cross-validation show that this FECNet achieves sensitivity 92.23 ± 2.14, specificity 93.18 0.87, precision 93.12 0.83, an accuracy 92.70 1.13 for first dataset, 92.19 1.89, 92.88 1.23, 92.83 1.22, 92.53 1.37 second dataset. We develop mobile app integrating model, web run cloud computing-based client–server modeled construction. This proposed corresponding effectively COVID-19, its performance better than five state-of-the-art recognition models.

Language: Английский

---

Deep Learning Framework for Liver Segmentation from T1-Weighted MRI Images

Sensors, Journal Year: 2023, Volume and Issue: 23(21), P. 8890 - 8890

Published: Nov. 1, 2023

The human liver exhibits variable characteristics and anatomical information, which is often ambiguous in radiological images. Machine learning can be of great assistance automatically segmenting the images, further processed for computer-aided diagnosis. Magnetic resonance imaging (MRI) preferred by clinicians pathology diagnosis over volumetric abdominal computerized tomography (CT) scans, due to their superior representation soft tissues. convenience Hounsfield unit (HoU) based preprocessing CT scans not available MRI, making automatic segmentation challenging MR This study investigates multiple state-of-the-art networks from MRI Here, T1-weighted (in-phase) are investigated using expert-labeled masks a public dataset 20 patients (647 slices) Combined Healthy Abdominal Organ Segmentation grant challenge (CHAOS). reason images that it demonstrates brighter fat content, thus providing enhanced task. Twenty-four different with varying depths dense, residual, inception encoder decoder backbones were A novel cascaded network proposed segment axial slices. framework outperforms existing approaches reported literature task (on same test set) dice similarity coefficient (DSC) score intersect union (IoU) 95.15% 92.10%, respectively.

Language: Английский

---

An elastic net regression model for predicting the risk of ICU admission and death for hospitalized patients with COVID-19

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: June 22, 2024

Abstract This study aimed to develop and validate prediction models estimate the risk of death intensive care unit admission in COVID-19 inpatients. All RT-PCR-confirmed adult inpatients admitted Fujian Provincial Hospital from October 2022 April 2023 were considered. Elastic Net Regression was used derive models. Potential factors considered, which included demographic characteristics, clinical symptoms, comorbidities, laboratory results, treatment process, prognosis. A total 1906 finally by inclusion/exclusion criteria divided into derivation test cohorts a ratio 8:2, where 1526 (80%) samples under repeated cross-validation framework remaining 380 (20%) for performance evaluation. Overall performance, discrimination calibration evaluated validation set cohort quantified accuracy, scaled Brier score (SbrS), area ROC curve (AUROC), Spiegelhalter-Z statistics. The performed well, with high levels (AUROC ICU [95%CI]: 0.858 [0.803,0.899]; AUROC 0.906 [0.850,0.948]); good calibrations (Spiegelhalter-Z : − 0.821 ( p -value: 0.412); 0.173) set. We developed validated help clinicians identify patients after infection.

Language: Английский

---

Machine learning-based prognostic model for 30-day mortality prediction in Sepsis-3

BMC Medical Informatics and Decision Making, Journal Year: 2024, Volume and Issue: 24(1)

Published: Sept. 9, 2024

Language: Английский

---

General Hospitalization and Intensive Care Unit-Related Factors of COVID-19 Patients in Northeastern Colombia: Baseline Characteristics of a Cohort Study

Cureus, Journal Year: 2023, Volume and Issue: unknown

Published: Aug. 21, 2023

Objective This study aims to describe demographic and clinical characteristics the factors associated with risk of COVID-19 general hospitalization intensive care unit (ICU) patients who consulted in a third-level hospital Santander, Colombia. Methods We used baseline data from an ambidirectional cohort study. included all positive real-time polymerase chain reaction (PCR) tests for came emergency room (ER) respiratory symptoms related COVID-19. Information regarding patients' was collected through telephone interviews review medical records. Vital signs were extracted records as well. Results enrolled 3,030 patients, predominantly men, median age 60 (interquartile range (IQR): 44-73). Symptoms acute phase varied between men women. Men presented more symptoms, women had symptoms. Hypertension, obesity, diabetes common admission. Antibiotic consumption may also play role Conclusions Male sex, older age, hypertension, prior thrombotic events, self-medicated antibiotics hospitalization. diabetes, cancer ICU The Charlson comorbidity index (CCI) is powerful tool evaluate impact pre-existing health conditions on highlight importance these findings possible predictors our region.

Language: Английский

---

Use of machine learning to identify protective factors for death from COVID-19 in the ICU: a retrospective study

PeerJ, Journal Year: 2024, Volume and Issue: 12, P. e17428 - e17428

Published: June 12, 2024

Background Patients in serious condition due to COVID-19 often require special care intensive units (ICUs). This disease has affected over 758 million people and resulted 6.8 deaths worldwide. Additionally, the progression of may vary from individual individual, that is, it is essential identify clinical parameters indicate a good prognosis for patient. Machine learning (ML) algorithms have been used analyzing complex medical data identifying prognostic indicators. However, there still an urgent need model elucidate predictors related patient outcomes. Therefore, this research aimed verify, through ML, variables involved discharge patients admitted ICU COVID-19. Methods In study, 126 were collected with information on demography, hospital length stay outcome, chronic diseases tumors, comorbidities risk factors, complications adverse events, health care, vital indicators southern Brazil. These filtered then selected by ML algorithm known as decision trees optimal set predicting using logistic regression. Finally, confusion matrix was performed evaluate model’s performance variables. Results Of 532 evaluated, 180 discharged: female (16.92%), central venous catheter (23.68%), bladder (26.13%), average 8.46- 23.65-days submitted mechanical ventilation, respectively. addition, chances increase 14% each additional day hospital, 136% patients, 716% when no catheter, 737% used. decrease 3% year age 9% other ventilation. The training presented balanced accuracy 0.81, sensitivity 0.74, specificity 0.88, kappa value 0.64. test had 0.85, 0.75, 0.95, 0.73. McNemar found significant differences error rates data, suggesting classification. work showed female, absence shorter duration associated greater chance discharge. results help develop measures lead

Language: Английский

---

Analysis and Application of Regression Models to ICU Patient Monitoring

Studies in computational intelligence, Journal Year: 2023, Volume and Issue: unknown, P. 301 - 318

Published: Jan. 1, 2023

Language: Английский

---