Comparing machine learning algorithms to predict COVID‑19 mortality using a dataset including chest computed tomography severity score data

Since the beginning of the COVID-19 pandemic, new and non-invasive digital technologies such as artificial intelligence (AI) had been introduced for mortality prediction of COVID-19 patients. The prognostic performances of the machine learning (ML)-based models for predicting clinical outcomes of COVID-19 patients had been mainly evaluated using demographics, risk factors, clinical manifestations, and laboratory results. There is a lack of information about the prognostic role of imaging manifestations in combination with demographics, clinical manifestations, and laboratory predictors. The purpose of the present study is to develop an efficient ML prognostic model based on a more comprehensive dataset including chest CT severity score (CT-SS). Fifty-five primary features in six main classes were retrospectively reviewed for 6854 suspected cases. The independence test of Chi-square was used to determine the most important features in the mortality prediction of COVID-19 patients. The most relevant predictors were used to train and test ML algorithms. The predictive models were developed using eight ML algorithms including the J48 decision tree (J48), support vector machine (SVM), multi-layer perceptron (MLP), k-nearest neighbourhood (k-NN), Naïve Bayes (NB), logistic regression (LR), random forest (RF), and eXtreme gradient boosting (XGBoost). The performances of the predictive models were evaluated using accuracy, precision, sensitivity, specificity, and area under the ROC curve (AUC) metrics. After applying the exclusion criteria, a total of 815 positive RT-PCR patients were the final sample size, where 54.85% of the patients were male and the mean age of the study population was 57.22 ± 16.76 years. The RF algorithm with an accuracy of 97.2%, the sensitivity of 100%, a precision of 94.8%, specificity of 94.5%, F1-score of 97.3%, and AUC of 99.9% had the best performance. Other ML algorithms with AUC ranging from 81.2 to 93.9% had also good prediction performances in predicting COVID-19 mortality. Results showed that timely and accurate risk stratification of COVID-19 patients could be performed using ML-based predictive models fed by routine data. The proposed algorithm with the more comprehensive dataset including CT-SS could efficiently predict the mortality of COVID-19 patients. This could lead to promptly targeting high-risk patients on admission, the optimal use of hospital resources, and an increased probability of survival of patients.

Diagnostic Efficacy of Chest Computed Tomography for Coronavirus Disease 2019

Background

A significant discrepancy between the results of previous studies is identified regarding the diagnostic efficacy of chest computed tomography (CT) for coronavirus disease 2019 (COVID-19). We aimed to evaluate the diagnostic efficacy of chest CT for COVID-19.

Methods

Suspected cases of COVID-19 with fever, cough, dyspnea, and evidence of pneumonia on chest CT scan were enrolled in the study. The accuracy, sensitivity, and specificity of chest CT were determined according to real-time reverse transcriptase-polymerase chain reaction (RT-PCR) results as the gold standard method.

Results

The study population comprised 356 suspected cases of COVID-19 (174 men and 182 women; age range 3-96 years; mean age ± standard deviation, 55.21 ± 18.38 years). COVID-19 patients were diagnosed using chest CT with 89.8% sensitivity, 78.1% accuracy, 21.3% specificity, 84.7% positive predictive value, and 30.23% negative predictive value. The odds ratio was 2.39 (95% confidence interval, 1.16-4.91). Typical CT manifestations of COVID-19 were observed in 48 (13.5%) patients with negative RT-PCR results and 30 (8.4%) patients with confirmed positive RT-PCR results had no radiological manifestations. Kappa coefficient of chest CT for diagnosis of COVID-19 was 0.78.

Conclusion

The results show that when RT-PCR results are negative, chest CT could be considered as a complementary diagnostic method for the diagnosis of COVID-19 patients. A more comprehensive diagnostic method could be established by combining the chest CT examination, clinical symptoms, and RT-PCR assay.

Prognostic significance of chest CT severity score in mortality prediction of COVID-19 patients, a machine learning study

Background

The high mortality rate of COVID-19 makes it necessary to seek early identification of high-risk patients with poor prognoses. Although the association between CT-SS and mortality of COVID-19 patients was reported, its prognosis significance in combination with other prognostic parameters was not evaluated yet.

Methods

This retrospective single-center study reviewed a total of 6854 suspected patients referred to Imam Khomeini hospital, Ilam city, west of Iran, from February 9, 2020 to December 20, 2020. The prognostic performances of k-Nearest Neighbors (kNN), Multilayer Perceptron (MLP), Support Vector Machine (SVM), and J48 decision tree algorithms were evaluated based on the most important and relevant predictors. The metrics derived from the confusion matrix were used to determine the performance of the ML models.

Results

After applying exclusion criteria, 815 hospitalized cases were entered into the study. Of these, 447(54.85%) were male and the mean (± SD) age of participants was 57.22(± 16.76) years. The results showed that the performances of the ML algorithms were improved when they are fed by the dataset with CT-SS data. The kNN model with an accuracy of 94.1%, sensitivity of 100. 0%, precision of 89.5%, specificity of 88.3%, and AUC around 97.2% had the best performance among the other three ML techniques.

Conclusions

The integration of CT-SS data with demographics, risk factors, clinical manifestations, and laboratory parameters improved the prognostic performances of the ML algorithms. An ML model with a comprehensive collection of predictors could identify high-risk patients more efficiently and lead to the optimal use of hospital resources.

Prognostic accuracy of visual lung damage computed tomography score for mortality prediction in patients with COVID-19 pneumonia: a systematic review and meta-analysis

Background

Chest computed tomography (CT) findings provide great added value in characterizing the extent of disease and severity of pulmonary involvements. Chest CT severity score (CT-SS) could be considered as an appropriate prognostic factor for mortality prediction in patients with COVID-19 pneumonia. In this study, we performed a meta-analysis evaluating the prognostic accuracy of CT-SS for mortality prediction in patients with COVID-19 pneumonia.

Methods

A systematic search was conducted on Web of Science, PubMed, Embase, Scopus, and Google Scholar databases between December 2019 and September 2021. The meta-analysis was performed using the random-effects model, and sensitivity and specificity (with 95%CIs) of CT-SS were calculated using the study authors’ pre-specified threshold.

Results

Sensitivity estimates ranged from 0.32 to 1.00, and the pooled estimate of sensitivity was 0.67 [95%CI (0.59–0.75)]. Specificity estimates ranged from 0.53 to 0.95 and the pooled estimate of specificity was 0.79 [95%CI (0.74–0.84)]. Results of meta-regression analysis showed that radiologist experiences did not affect the sensitivity and specificity of CT-SS to predict mortality in COVID-19 patients (P = 0.314 and 0.283, respectively). The test for subgroup differences suggests that study location significantly modifies sensitivity and specificity of CT-SS to predict mortality in COVID-19 patients. The area under the summary receiver operator characteristic (ROC) curve was 0.8248.

Conclusion

Our results have shown that CT-SS has acceptable prognostic accuracy for mortality prediction in COVID-19 patients. This simple scoring method could help to improve the management of high-risk patients with COVID-19.

Association of chest CT severity score with mortality of COVID-19 patients: a systematic review and meta-analysis

Purpose

Chest computed tomography (CT) is a high-sensitivity diagnostic tool for depicting interstitial pneumonia and may lay a critical role in the evaluation of the severity and extent of pulmonary involvement. In this study, we aimed to evaluate the association of chest CT severity score (CT-SS) with the mortality of COVID-19 patients using systematic review and meta-analysis.

Methods

Web of Science, PubMed, Embase, Scopus, and Google Scholar were used to search for primary articles. The meta-analysis was performed using the random-effects model, and odds ratios (ORs) with 95% confidence intervals (95%CIs) were calculated as the effect sizes.

Results

This meta-analysis retrieved a total number of 7106 COVID-19 patients. The pooled estimate for the association of CT-SS with mortality of COVID-19 patients was calculated as 1.244 (95% CI 1.157–1.337). The pooled estimate for the association of CT-SS with an optimal cutoff and mortality of COVID-19 patients was calculated as 7.124 (95% CI 5.307–9.563). There was no publication bias in the results of included studies. Radiologist experiences and study locations were not potential sources of between-study heterogeneity (both P > 0.2). The shapes of Begg’s funnel plots seemed symmetrical for studies evaluating the association of CT-SS with/without the optimal cutoffs and mortality of COVID-19 patients (Begg’s test P = 0.945 and 0.356, respectively).

Conclusions

The results of this study point to an association between CT-SS and mortality of COVID-19 patients. The odds of mortality for COVID-19 patients could be accurately predicted using an optimal CT-SS cutoff in visual scoring of lung involvement.

Sensitivity Laplacian Ratio-Based Optimization of the Projection Selection for Diffuse Optical Tomography

BACKGROUND

In diffuse optical tomography, determining the optimal angle between the source and detector is an effective method to reduce the number of projections while maintaining the quality of the reconstructed images. In this study, a new parameter is introduced to evaluate the source-detector geometries.

METHODS

A two-dimensional mesh with the radius of 20 mm and 7987 nodes were built. In each reconstruction, 0.5 mm heterogeneity with the absorption coefficient of 0.06 mm^-1 and the dispersion coefficient of 0.6 mm^-1 was added in different parts of the sample randomly. The relationship between the mean square error (MSE), sensitivity Laplacian ratio (SLR), and sensitivity standard deviation ratio (SSR) was evaluated based on their correlation coefficients. The quality of the images achieved using the optimized projections were compared with that of the full projections for the same depths.

RESULTS

MSE decreases by increasing the SLR magnitudes which indicate that the parameter could be used to evaluate the scanning geometries. There was a negative correlation coefficient (R = -0.76) with the inverse relationship between the SLR and MSE indices. SSR does not have a significant relationship with the quality of the reconstructed images. For each scanning depth, the comparison of the images obtained using the full and optimized-selective projections did not show any considerable difference despite the decrease of the projection numbers in scanning geometry with the optimized-selective projections.

CONCLUSION

The unnecessary projections could be eliminated by placing the detectors at the specific angles, which were determined using the SLR. Thus, a proper compromise between the quality of the reconstructed images and reconstruction time might establish.

Sensitivity Uniformity Ratio as a New Index to Optimize the Scanning Geometry for Fluorescent Molecular Tomography

Background:

Molecular fluorescence imaging is widely used as a noninvasive method to study the cellular and molecular mechanisms. In the optical imaging system, the sensitivity is the change of the intensity received by the detector for the changed optical characteristics (fluorescence) in each sample point. Sensitivity could be considered as a function of imaging geometry. A favor imaging system has a uniform and high-sensitivity coefficient for each point of the sample. In this study, a new parameter was proposed which the optimal angle between the source and detector could be determined based on this parameter.

Methods:

For evaluation of the new method, a two-dimensional mesh with a radius of 20 mm and 5133 nodes was built. In each reconstruction, 0.5-mm fluorescence heterogeneity with a contrast-to-purpose ratio of fluorescence yield of 10 was randomly added at different points of the sample. The source and the detector were simulated in different geometric conditions. The calculations were performed using the NIRFAST and MATLAB software. The relationship between mean squared error (MSE) and sensitivity uniformity ratio (SUR) was evaluated using the correlation coefficient. Finally, based on the new index, an optimal geometrical strategy was introduced.

Results:

There was a negative correlation coefficient (R = −0.78) with the inverse relationship between the SUR and MSE indices. The reconstructed images showed that the better image quality achieved using the optimal geometry for all scanning depths. For the conventional geometry, there is an artifact in the opposite side of the inhomogeneity at the shallow depths, which has been eliminated in the reconstructed images achieved using the optimal geometry.

Conclusion:

The SUR is a powerful computational tool which could be used to determine the optimal angles between the source and detector for each scanning depth.

Physical, dosimetric and clinical aspects and delivery systems in neutron capture therapy

Neutron capture therapy (NCT) is a targeted radiotherapy for cancer treatment. In this method, neutrons with a spectra/specific energy (depending on the type of agent used for NCT) are captured with an agent that has a high cross-section with these neutrons. There are some agents that have been proposed in NCT including 10B, 157Gd and 33S. Among these agents, only 10B is used in clinical trials. Application of 157Gd is limited to in-vivo and in-vitro research. In addition, 33S has been applied in the field of Monte Carlo simulation. In BNCT, the only two delivery agents which are presently applied in clinical trials are BPA and BSH, but other delivery systems are being developed for more effective treatment in NCT. Neutron sources used in NCT are fission reactors, accelerators, and 252Cf. Among these, fission reactors have the most application in NCT. So far, BNCT has been applied to treat various cancers including glioblastoma multiforme, malignant glioma, malignant meningioma, liver, head and neck, lung, colon, melanoma, thyroid, hepatic, gastrointestinal cancer, and extra-mammary Paget's disease. This paper aims to review physical, dosimetric and clinical aspects as well as delivery systems in NCT for various agents.

Assessment of the Agreement between Cerebral Hemodynamic Indices Quantified Using Dynamic Susceptibility Contrast and Dynamic Contrast-enhanced Perfusion Magnetic Resonance Imagings

Background:

Brain tumor is one of the most common tumors. A successful treatment might be achieved with an early identification. Pathological investigation as the gold standard method for tumor identification has some limitations. Noninvasive assessment of tumor specifications may be possible using perfusion-weighted magnetic resonance imaging (MRI). Cerebral blood volume (CBV) and cerebral blood flow (CBF) could be calculated based on dynamic contrast-enhanced MRI (DCE-MRI) in addition to dynamic susceptibility contrast MRI (DSC-MRI) modality. Each category of the cerebral hemodynamic and permeability indices revealed the specific tumor characteristics and their collection could help for better identification of the tumor. Some mathematical methods were developed to determine both cerebral hemodynamic and permeability indices based on a single-dose DCE perfusion MRI. There are only a few studies available on the comparison of DSC- and DCE-derived cerebral hemodynamic indices such as CBF and CBV.

Aim:

The objective of the study was to validate first-pass perfusion parameters derived from T1-based DCE method in comparison to the routine T2*-based DSC protocol.

Materials and Methods:

Twenty-nine patients with brain tumor underwent DCE- and DSC-MRIs to evaluate the agreement between DSC- and DCE-derived cerebral hemodynamic parameters. Agreement between DSC- and DCE-derived cerebral hemodynamic indices was determined using the statistical method described by Bland and Altman. The reliability between DSC- and DCE-derived cerebral hemodynamic indices was measured using the intraclass correlation analysis.

Results:

The achieved magnitudes for DCE-derived CBV (gray matter [GM]: 5.01 ± 1.40 mL/100 g vs. white matter [WM]: 1.84 ± 0.74 mL/100 g) and DCE-derived CBF (GM: 60.53 ± 12.70 mL/100 g/min vs. WM: 32.00 ± 6.00 mL/100 g/min) were in good agreement with other studies. The intraclass correlation coefficients showed that the cerebral hemodynamic indices could accurately be estimated based on the DCE-MRI using a single-compartment model (>0.87), and DCE-derived cerebral hemodynamic indices are significantly similar to the magnitudes achieved based on the DSC-MRI (P < 0.001). Furthermore, an acceptable agreement was observed between DSC- and DCE-derived cerebral hemodynamic indices.

Conclusion:

Based on the measurement of the cerebral hemodynamic and blood–brain barrier permeability using DCE-MRI, a more comprehensive collection of the physiological parameters cloud be achieved for tumor evaluations.

Gastrointestinal Side Effects of the Radioiodine Therapy for the Patients with Differentiated Thyroid Carcinoma Two Days after Prescription

Iodine-131 (I-131) therapy is one of the conventional approaches in the treatment of patients with differentiated thyroid carcinoma (DTC). The radioiodine agents also accumulate in the other organs that cause pain and damage to the patients. Radioiodine therapy is associated with various gastrointestinal (GI) toxicities. In this study, GI side effects of the radioiodine therapy were investigated. GI toxicities of the radioiodine therapy were studied in 137 patients with histologically proven DTC in Jun-Nov 2014. All the patients were treated by radioiodine agents in the research institute of Shariati Hospital, Tehran, Iran. The patients were examined 48 h after prescription (before discharge) and their GI side effects were registered. Correlation of the age, gender, administered dose, administered dose per body weight as the independent factors, and GI side effects were analyzed using the Pearson correlation test with Statistical Package for the Social Sciences (SPSS) version 20. Regression coefficients and linearity of the variable were investigated by MATLAB software. Line fitting was performed using MATLAB curve-fitting toolbox. From the subjects, 38 patients had GI complaints (30.4%). Significant factors influencing GI side effects were dose per body weight and administered doses. There was no significant correlation between age and gender as the independent parameters and GI complaints. The most prevalent GI side effect was nausea that occurs in 26.4% of the patients. From the results, it could be concluded that the GI side effects could be prevented by administering a safe radioiodine dose value less than 5,550 MBq.

The Effect of the Lumbar Vertebral Malpositioning on Bone Mineral Density Measurements of the Lumbar Spine by Dual-Energy X-Ray Absorptiometry

A significant discrepancy between the results of previous human and phantoms studies is identified regarding the effects of vertebral positioning on bone mineral density (BMD) measurements. We aimed to evaluate the effects of lumbar vertebral positioning on BMD measurements by dual-energy X-ray absorptiometry in a human cadaveric spine phantom. A spine phantom was designed using L1-L4 vertebrae harvested from a 48-year-old male cadaver without coronal or sagittal deformity. The spine phantom was scanned by DEXXUM T bone densitometer in a constant scanning speed of 30 mm/s and resolution of 1.0 × 1.0 mm. BMD values were measured in a positive and negative lumbar lordosis and kyphosis tilt angles in the sagittal plane, from 0° to 35°, with 7° increments. Also BMD values were measured in axial and lateral rotations with 5° increments. Projectional dual-energy X-ray absorptiometry measurements are significantly affected by positioning of the lumbar spine, more severely affected by kyphotic curvature, but also by axial and lateral rotational scoliosis as well as lordotic curvature. Increasing the severity of lordosis and kyphosis curvatures leads to false reduction of BMD value up to 17.5% and 11.5%, respectively. Increasing the degree of lateral and axial rotational scolioses results in a false decrease in BMD measurements by up to 10.8% and 9.6%, respectively. To achieve the most accurate scanning results, error sources and abnormal positioning should be identified and minimized as much as possible. If not correctable, they should be taken into consideration while interpreting the results.

Effective parameters on conductivity of mineralized carbon nanofibers: an investigation using artificial neural networks

The aim of this study was to predict the effects of different parameters on the conductivity of mineralized PAN-based carbon nanofibers by the artificial neural network (ANN) method.

Assessment of target volume doses in radiotherapy based on the standard and measured calibration curves

CONTEXT

In radiation treatments, estimation of the dose distribution in the target volume is one of the main components of the treatment planning procedure. To estimate the dose distribution, the information of electron densities is necessary. The standard curves determined by computed tomography (CT) scanner that may be different from that of other oncology centers. In this study, the changes of dose calculation due to the different calibration curves (HU-ρel) were investigated.

MATERIALS AND METHODS

Dose values were calculated based on the standard calibration curve that was predefined for the treatment planning system (TPS). The calibration curve was also extracted from the CT images of the phantom, and dose values were calculated based on this curve. The percentage errors of the calculated values were determined.

STATISTICAL ANALYSIS USED

The statistical analyses of the mean differences were performed using the Wilcoxon rank-sum test for both of the calibration curves.

RESULTS AND DISCUSSION

The results show no significant difference for both of the measured and standard calibration curves (HU-ρel) in 6, 15, and 18 MeV energies. In Wilcoxon ranked sum nonparametric test for independent samples with P<0.05, the equality of monitor units for both of the curves to transfer 200 cGy doses to reference points was resulted. The percentage errors of the calculated values were lower than 2% and 1.5% in 6 and 15 MeV, respectively.

CONCLUSION

From the results, it could be concluded that the standard calibration curve could be used in TPS dose calculation accurately.

Validation of a Prototype Optical Computed Tomography System

In radiation cancer treatments, the most of the side effects could be minimized using a proper dosimeter. Gel dosimeter is the only three-dimensional dosimeter and magnetic resonance imaging (MRI) is the gold standard method for gel dosimeter readout. Because of hard accessibility and high cost of sample reading by MRI systems, some other alternative methods were developed. The optical computed tomography (OCT) method could be considered as the most promising alternative method that has been studied widely. In the current study, gel dosimeter scanning using a prototype optical scanner and validation of this optical scanner was performed. Optical absorbance of the irradiated gel samples was determined by both of conventional spectrophotometer and the fabricated OCT system at 632 nm. Furthermore, these irradiated vials were scanned by a 1.5 T MRI. The slope of the curves was extracted as the dose-response sensitivity. The R2-dose sensitivity measured by MRI method was 0.1904 and 0.113 for NIPAM and PAGAT gels, respectively. The optical dose sensitivity obtained by conventional spectrophotometer and the fabricated optical scanner was 0.0453 and 0.0442 for NIPAM gels and 0.0244 and 0.0242 for PAGAT gels, respectively. The scanning results of the absorbed dose values showed that the new OCT and conventional spectrophotometer were in fair agreement. From the results, it could be concluded that the fabricated system is able to quantize the absorbed dose values in polymer gel samples with acceptable accuracy.

Dosimetric optimization of a colpostat in a 60Co high-dose-rate brachytherapy unit for bladder sparing

BACKGROUND

Cervical cancer brachytherapy has an effective role on the tumor control probability as a boost and/or single treatment option. High-dose-rate (60)Co brachytherapy units are used in many radiation oncology centers in Iran. Rectum and bladder tissues are considered as organs at risk in radiation treatment of cervical cancers, and they should be spared from unwanted radiation risk.

PURPOSE

In the present study, the effect of additional tungsten shield was investigated in a new colpostat design for bladder protection against radiation.

METHODS AND MATERIALS

Monte Carlo (MC) simulation has been performed using the MC N-Particle eXtended version 2.4.0 transport code. The HDR GZP6 brachytherapy source applicator was simulated along with its colpostats. The f6 tally was used for absorbed dose calculation; and for model validation, we used from dosimetric features of the GZP6 treatment planning system.

RESULTS

Results calculated by MC simulation method showed that dose reduction at the end of the colpostat was 2.44% for the medium colpostat and a dose increase of 1.35% was obtained for the small colpostat. In the reference point of the bladder (at the distance 1cm from the end point of the colpostat), the percentages of dose reductions were also 25% and 15% for medium and small colpostats, respectively.

CONCLUSIONS

Results show that the absorbed dose in the bladder tissue can be reduced significantly using a shielded colpostat.

Design and construction of an optical computed tomography scanner for polymer gel dosimetry application

Polymer gel dosimeter is the only accurate three dimensional (3D) dosimeter that can measure the absorbed dose distribution in a perfect 3D setting. Gel dosimetry by using optical computed tomography (OCT) has been promoted by several researches. In the current study, we designed and constructed a prototype OCT system for gel dosimetry. First, the electrical system for optical scanning of the gel container using a Helium-Neon laser and a photocell was designed and constructed. Then, the mechanical part for two rotational and translational motions was designed and step motors were assembled to it. The data coming from photocell was grabbed by the home-built interface and sent to a personal computer. Data processing was carried out using MATLAB software. To calibrate the system and tune up the functionality of it, different objects was designed and scanned. Furthermore, the spatial and contrast resolution of the system was determined. The system was able to scan the gel dosimeter container with a diameter up to 11 cm inside the water phantom. The standard deviation of the pixels within water flask image was considered as the criteria for image uniformity. The uniformity of the system was about ±0.05%. The spatial resolution of the system was approximately 1 mm and contrast resolution was about 0.2%. Our primary results showed that this system is able to obtain two-dimensional, cross-sectional images from polymer gel samples.