Deep Learning With Chest Radiographs for Making Prognoses in Patients With COVID-19: Retrospective Cohort Study

The Application of Mask Region-Based Convolutional Neural Networks in the Detection of Nasal Septal Deviation Using Cone Beam Computed Tomography Images: Proof-of-Concept Study

BACKGROUND

Artificial intelligence (AI) models are being increasingly studied for the detection of variations and pathologies in different imaging modalities. Nasal septal deviation (NSD) is an important anatomical structure with clinical implications. However, AI-based radiographic detection of NSD has not yet been studied.

OBJECTIVE

This research aimed to develop and evaluate a real-time model that can detect probable NSD using cone beam computed tomography (CBCT) images.

METHODS

Coronal section images were obtained from 204 full-volume CBCT scans. The scans were classified as normal and deviated by 2 maxillofacial radiologists. The images were then used to train and test the AI model. Mask region-based convolutional neural networks (Mask R-CNNs) comprising 3 different backbones-ResNet50, ResNet101, and MobileNet-were used to detect deviated nasal septum in 204 CBCT images. To further improve the detection, an image preprocessing technique (contrast enhancement [CEH]) was added.

RESULTS

The best-performing model-CEH-ResNet101-achieved a mean average precision of 0.911, with an area under the curve of 0.921.

CONCLUSIONS

The performance of the model shows that the model is capable of detecting nasal septal deviation. Future research in this field should focus on additional preprocessing of images and detection of NSD based on multiple planes using 3D images.

[The contribution of artificial intelligence (AI) subsequent to the processing of thoracic imaging]

The contribution of artificial intelligence (AI) to medical imaging is currently the object of widespread experimentation. The development of deep learning (DL) methods, particularly convolution neural networks (CNNs), has led to performance gains often superior to those achieved by conventional methods such as machine learning. Radiomics is an approach aimed at extracting quantitative data not accessible to the human eye from images expressing a disease. The data subsequently feed machine learning models and produce diagnostic or prognostic probabilities. As for the multiple applications of AI methods in thoracic imaging, they are undergoing evaluation. Chest radiography is a practically ideal field for the development of DL algorithms able to automatically interpret X-rays. Current algorithms can detect up to 14 different abnormalities present either in isolation or in combination. Chest CT is another area offering numerous AI applications. Various algorithms have been specifically formed and validated for the detection and characterization of pulmonary nodules and pulmonary embolism, as well as segmentation and quantitative analysis of the extent of diffuse lung diseases (emphysema, infectious pneumonias, interstitial lung disease). In addition, the analysis of medical images can be associated with clinical, biological, and functional data (multi-omics analysis), the objective being to construct predictive approaches regarding disease prognosis and response to treatment.

Effectiveness of COVID-19 Vaccines Over Time Against Clinical and Radiologic Outcomes Related to Severe SARS-CoV-2 Infection

Background The impact of waning vaccine effectiveness on the severity of COVID-19-related findings discovered with radiologic examinations remains underexplored. Purpose To evaluate the effectiveness of vaccines over time against severe clinical and radiologic outcomes related to SARS-CoV-2 infections. Materials and Methods This multicenter retrospective study included patients in the Korean Imaging Cohort of COVID-19 database who were hospitalized for COVID-19 between June 2021 and December 2022. Patients who had received at least one dose of a SARS-CoV-2 vaccine were categorized based on the time elapsed between diagnosis and their last vaccination. Adjusted multivariable logistic regression analysis was used to estimate vaccine effectiveness against a composite of severe clinical outcomes (invasive ventilation, extracorporeal membrane oxygenation, or in-hospital death) and severe radiologic pneumonia (≥25% of lung involvement), and odds ratios (ORs) were compared between patients vaccinated within 90 days of diagnosis and those vaccinated more than 90 days before diagnosis. Results Of 4196 patients with COVID-19 (mean age, 66 years ± 17 [SD]; 2132 [51%] women, 2064 [49%] men), the ratio of severe pneumonia since their most recent vaccination was as follows: 90 days or less, 18% (277 of 1527); between 91 and 120 days, 22% (172 of 783); between 121 and 180 days, 27% (274 of 1032); between 181 and 240 days, 32% (159 of 496); and more than 240 days, 31% (110 of 358). Patients vaccinated more than 240 days before diagnosis showed increased odds of severe clinical outcomes compared with patients vaccinated within 90 days (OR = 1.94 [95% CI: 1.16, 3.24]; P = .01). Similarly, patients vaccinated more than 240 days before diagnosis showed increased odds of severe pneumonia on chest radiographs compared with patients vaccinated within 90 days (OR = 1.65 [95% CI: 1.13, 2.40]; P = .009). No difference in odds of severe clinical outcomes (P = .13 to P = .68) or severe pneumonia (P = .15 to P = .86) were observed between patients vaccinated 91-240 days before diagnosis and those vaccinated within 90 days of diagnosis. Conclusion Vaccine effectiveness against severe clinical outcomes and severe pneumonia related to SARS-CoV-2 infection gradually declined, with increased odds of both observed in patients vaccinated more than 240 days before diagnosis. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Wells in this issue.

A deep learning model for predicting COVID-19 ARDS in critically ill patients

Background

The coronavirus disease 2019 (COVID-19) is an acute infectious pneumonia caused by a severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection previously unknown to humans. However, predictive studies of acute respiratory distress syndrome (ARDS) in patients with COVID-19 are limited. In this study, we attempted to establish predictive models to predict ARDS caused by COVID-19 via a thorough analysis of patients' clinical data and CT images.

Method

The data of included patients were retrospectively collected from the intensive care unit in our hospital from April 2022 to June 2022. The primary outcome was the development of ARDS after ICU admission. We first established two individual predictive models based on extreme gradient boosting (XGBoost) and convolutional neural network (CNN), respectively; then, an integrated model was developed by combining the two individual models. The performance of all the predictive models was evaluated using the area under receiver operating characteristic curve (AUC), confusion matrix, and calibration plot.

Results

A total of 103 critically ill COVID-19 patients were included in this research, of which 23 patients (22.3%) developed ARDS after admission; five predictive variables were selected and further used to establish the machine learning models, and the XGBoost model yielded the most accurate predictions with the highest AUC (0.94, 95% CI: 0.91-0.96). The AUC of the CT-based convolutional neural network predictive model and the integrated model was 0.96 (95% CI: 0.93-0.98) and 0.97 (95% CI: 0.95-0.99), respectively.

Conclusion

An integrated deep learning model could be used to predict COVID-19 ARDS in critically ill patients.

Unraveling complex relationships between COVID-19 risk factors using machine learning based models for predicting mortality of hospitalized patients and identification of high-risk group: a large retrospective study

Background

At the end of 2019, the coronavirus disease 2019 (COVID-19) pandemic increased the hospital burden of COVID-19 caused by the SARS-Cov-2 and became the most significant health challenge for nations worldwide. The severity and high mortality of COVID-19 have been correlated with various demographic characteristics and clinical manifestations. Prediction of mortality rate, identification of risk factors, and classification of patients played a crucial role in managing COVID-19 patients. Our purpose was to develop machine learning (ML)-based models for the prediction of mortality and severity among patients with COVID-19. Identifying the most important predictors and unraveling their relationships by classification of patients to the low-, moderate- and high-risk groups might guide prioritizing treatment decisions and a better understanding of interactions between factors. A detailed evaluation of patient data is believed to be important since COVID-19 resurgence is underway in many countries.

Results

The findings of this study revealed that the ML-based statistically inspired modification of the partial least square (SIMPLS) method could predict the in-hospital mortality among COVID-19 patients. The prediction model was developed using 19 predictors including clinical variables, comorbidities, and blood markers with moderate predictability (Q² = 0.24) to separate survivors and non-survivors. Oxygen saturation level, loss of consciousness, and chronic kidney disease (CKD) were the top mortality predictors. Correlation analysis showed different correlation patterns among predictors for each non-survivor and survivor cohort separately. The main prediction model was verified using other ML-based analyses with a high area under the curve (AUC) (0.81-0.93) and specificity (0.94-0.99). The obtained data revealed that the mortality prediction model can be different for males and females with diverse predictors. Patients were classified into four clusters of mortality risk and identified the patients at the highest risk of mortality, which accentuated the most significant predictors correlating with mortality.

Conclusion

An ML model for predicting mortality among hospitalized COVID-19 patients was developed considering the interactions between factors that may reduce the complexity of clinical decision-making processes. The most predictive factors related to patient mortality were identified by assessing and classifying patients into different groups based on their sex and mortality risk (low-, moderate-, and high-risk groups).