ABSTRACT
ObjectiveThe current pandemic has led to a proliferation of predictive models being developed to address various aspects of COVID-19 patient care. We aimed to develop an online platform that would serve as an open source repository for a curated subset of such models, and provide a simple interface for included models to allow for online calculation. This platform would support doctors during decision-making regarding diagnoses, prognoses, and follow-up of COVID-19 patients, expediting the models transition from research to clinical practice. MethodsIn this proof-of-principle study, we performed a literature search in PubMed and WHO database to find suitable models for implementation on our platform. All selected models were publicly available (peer reviewed publications or open source repository) and had been validated (TRIPOD type 3 or 2b). We created a method for obtaining the regression coefficients if only the nomogram was available in the original publication. All predictive models were transcribed on a practical graphical user interface using PHP 8.0.0, and published online together with supporting documentation and links to the associated articles. ResultsThe open source website https://covid19risk.ai/ currently incorporates nine models from six different research groups, evaluated on datasets from different countries. The website will continue to be populated with other models related to COVID-19 prediction as these become available. This dynamic platform allows COVID-19 researchers to contact us to have their model curated and included on our website, thereby increasing the reach and real-world impact of their work. ConclusionWe have successfully demonstrated in this proof-of-principle study that our website provides an inclusive platform for predictive models related to COVID-19. It enables doctors to supplement their judgment with patient-specific predictions from externally-validated models in a user-friendly format. Additionally, this platform supports researchers in showcasing their work, which will increase the visibility and use of their models.
ABSTRACT
Objective@#To investigate the image quality of ultralow-dose CT (ULDCT) of the chest reconstructed using a cycle-consistent generative adversarial network (CycleGAN)-based deep learning method in the evaluation of pulmonary tuberculosis. @*Materials and Methods@#Between June 2019 and November 2019, 103 patients (mean age, 40.8 ± 13.6 years; 61 men and 42 women) with pulmonary tuberculosis were prospectively enrolled to undergo standard-dose CT (120 kVp with automated exposure control), followed immediately by ULDCT (80 kVp and 10 mAs). The images of the two successive scans were used to train the CycleGAN framework for image-to-image translation. The denoising efficacy of the CycleGAN algorithm was compared with that of hybrid and model-based iterative reconstruction. Repeated-measures analysis of variance and Wilcoxon signedrank test were performed to compare the objective measurements and the subjective image quality scores, respectively. @*Results@#With the optimized CycleGAN denoising model, using the ULDCT images as input, the peak signal-to-noise ratio and structural similarity index improved by 2.0 dB and 0.21, respectively. The CycleGAN-generated denoised ULDCT images typically provided satisfactory image quality for optimal visibility of anatomic structures and pathological findings, with a lower level of image noise (mean ± standard deviation [SD], 19.5 ± 3.0 Hounsfield unit [HU]) than that of the hybrid (66.3 ± 10.5 HU, p 0.908). The CycleGAN-generated images showed the highest contrast-to-noise ratios for the pulmonary lesions, followed by the model-based and hybrid iterative reconstruction. The mean effective radiation dose of ULDCT was 0.12 mSv with a mean 93.9% reduction compared to standard-dose CT. @*Conclusion@#The optimized CycleGAN technique may allow the synthesis of diagnostically acceptable images from ULDCT of the chest for the evaluation of pulmonary tuberculosis.
ABSTRACT
Objective@#To investigate the image quality of ultralow-dose CT (ULDCT) of the chest reconstructed using a cycle-consistent generative adversarial network (CycleGAN)-based deep learning method in the evaluation of pulmonary tuberculosis. @*Materials and Methods@#Between June 2019 and November 2019, 103 patients (mean age, 40.8 ± 13.6 years; 61 men and 42 women) with pulmonary tuberculosis were prospectively enrolled to undergo standard-dose CT (120 kVp with automated exposure control), followed immediately by ULDCT (80 kVp and 10 mAs). The images of the two successive scans were used to train the CycleGAN framework for image-to-image translation. The denoising efficacy of the CycleGAN algorithm was compared with that of hybrid and model-based iterative reconstruction. Repeated-measures analysis of variance and Wilcoxon signedrank test were performed to compare the objective measurements and the subjective image quality scores, respectively. @*Results@#With the optimized CycleGAN denoising model, using the ULDCT images as input, the peak signal-to-noise ratio and structural similarity index improved by 2.0 dB and 0.21, respectively. The CycleGAN-generated denoised ULDCT images typically provided satisfactory image quality for optimal visibility of anatomic structures and pathological findings, with a lower level of image noise (mean ± standard deviation [SD], 19.5 ± 3.0 Hounsfield unit [HU]) than that of the hybrid (66.3 ± 10.5 HU, p 0.908). The CycleGAN-generated images showed the highest contrast-to-noise ratios for the pulmonary lesions, followed by the model-based and hybrid iterative reconstruction. The mean effective radiation dose of ULDCT was 0.12 mSv with a mean 93.9% reduction compared to standard-dose CT. @*Conclusion@#The optimized CycleGAN technique may allow the synthesis of diagnostically acceptable images from ULDCT of the chest for the evaluation of pulmonary tuberculosis.
ABSTRACT
Key pointsO_ST_ABSQuestionC_ST_ABSHow do nomograms and machine-learning algorithms of severity risk prediction and triage of COVID-19 patients at hospital admission perform? FindingsThis model was prospectively validated on six test datasets comprising of 426 patients and yielded AUCs ranging from 0.816 to 0.976, accuracies ranging from 70.8% to 93.8%, sensitivities ranging from 83.7% to 100%, and specificities ranging from 41.0% to 95.7%. The cut-off probability values for low, medium, and high-risk groups were 0.072 and 0.244. MeaningThe findings of this study suggest that our models performs well for the diagnosis and prediction of progression to severe or critical illness of COVID-19 patients and could be used for triage of COVID-19 patients at hospital admission. IMPORTANCEThe outbreak of the coronavirus disease 2019 (COVID-19) has globally strained medical resources and caused significant mortality for severely and critically ill patients. However, the availability of validated nomograms and the machine-learning model to predict severity risk and triage of affected patients is limited. OBJECTIVETo develop and validate nomograms and machine-learning models for severity risk assessment and triage for COVID-19 patients at hospital admission. DESIGN, SETTING, AND PARTICIPANTSA retrospective cohort of 299 consecutively hospitalized COVID-19 patients at The Central Hospital of Wuhan, China, from December 23, 2019, to February 13, 2020, was used to train and validate the models. Six cohorts with 426 patients from eight centers in China, Italy, and Belgium, from February 20, 2020, to March 21, 2020, were used to prospectively validate the models. MAIN OUTCOME AND MEASURESThe main outcome was the onset of severe or critical illness during hospitalization. Model performances were quantified using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity. RESULTSOf the 299 hospitalized COVID-19 patients in the retrospective cohort, the median age was 50 years ((interquartile range, 35.5-63.0; range, 20-94 years) and 137 (45.8%) were men. Of the 426 hospitalized COVID-19 patients in the prospective cohorts, the median age was 62.0 years ((interquartile range, 50.0-72.0; range, 19-94 years) and 236 (55.4%) were men. The model was prospectively validated on six cohorts yielding AUCs ranging from 0.816 to 0.976, with accuracies ranging from 70.8% to 93.8%, sensitivities ranging from 83.7% to 100%, and specificities ranging from 41.0% to 95.7%. The cut-off values of the low, medium, and high-risk probabilities were 0.072 and 0.244. The developed online calculators can be found at https://covid19risk.ai/. CONCLUSION AND RELEVANCEThe machine learning models, nomograms, and online calculators might be useful for the prediction of onset of severe and critical illness among COVID-19 patients and triage at hospital admission. Further prospective research and clinical feedback are necessary to evaluate the clinical usefulness of this model and to determine whether these models can help optimize medical resources and reduce mortality rates compared with current clinical practices.
ABSTRACT
BackgroundThe coronavirus disease 2019 (COVID-19) outbreak has reached pandemic status. Drastic measures of social distancing are enforced in society and healthcare systems are being pushed to and over their limits. ObjectivesTo develop a fully automatic framework to detect COVID-19 by applying AI to chest CT and evaluate validation performance. MethodsIn this retrospective multi-site study, a fully automated AI framework was developed to extract radiomics features from volumetric chest CT exams to learn the detection pattern of COVID-19 patients. We analysed the data from 181 RT-PCR confirmed COVID-19 patients as well as 1200 other non-COVID-19 control patients to build and assess the performance of the model. The datasets were collected from 2 different hospital sites of the CHU Liege, Belgium. Diagnostic performance was assessed by the area under the receiver operating characteristic curve (AUC), sensitivity and specificity. Results1381 patients were included in this study. The average age was 64.4{+/-}15.8 and 63.8{+/-}14.4 years with a gender balance of 56% and 52% male in the COVID-19 and control group, respectively. The final curated dataset used for model construction and validation consisted of chest CT scans of 892 patients. The model sensitivity and specificity for detecting COVID-19 in the test set (training 80% and test 20% of patients) were 78.94% and 91.09%, respectively, with an AUC of 0.9398 (95% CI: 0.875-1). The negative predictive value of the algorithm was found to be larger than 97%. ConclusionsBenchmarked against RT-PCR confirmed cases of COVID-19, our AI framework can accurately differentiate COVID-19 from routine clinical conditions in a fully automated fashion. Thus, providing rapid accurate diagnosis in patients suspected of COVID-19 infection, facilitating the timely implementation of isolation procedures and early intervention.
