Development and Validation of an Automated Radiomic CT Signature for Detecting COVID-19

The 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 beyond their limits. To help in the fight against this threat on human health, a fully automated AI framework was developed to extract radiomics features from volumetric chest computed tomography (CT) exams. The detection model was developed on a dataset of 1381 patients (181 COVID-19 patients plus 1200 non COVID control patients). A second, independent dataset of 197 RT-PCR confirmed COVID-19 patients and 500 control patients was used to assess the performance of the model. Diagnostic performance was assessed by the area under the receiver operating characteristic curve (AUC). The model had an AUC of 0.882 (95% CI: 0.851–0.913) in the independent test dataset (641 patients). The optimal decision threshold, considering the cost of false negatives twice as high as the cost of false positives, resulted in an accuracy of 85.18%, a sensitivity of 69.52%, a specificity of 91.63%, a negative predictive value (NPV) of 94.46% and a positive predictive value (PPV) of 59.44%. Benchmarked 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.

Download Full-text

Development and validation of an automated radiomic CT signature for detecting COVID-19

10.1101/2020.04.28.20082966 ◽

2020 ◽

Cited By ~ 1

Author(s):

J. Guiot ◽

A. Vaidyanathan ◽

L. Deprez ◽

F. Zerka ◽

L. Danthine ◽

...

Keyword(s):

Sensitivity And Specificity ◽

Operating Characteristic ◽

Characteristic Curve ◽

Chest Ct ◽

Control Group ◽

Rt Pcr ◽

Fully Automatic ◽

Clinical Conditions ◽

Development And Validation ◽

And Control

AbstractBackgroundThe 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 Liège, 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.

Download Full-text

Triage of Patients Suspected of COVID-19 in Chronic Hemodialysis: Eosinophil Count Differentiates Low and High Suspicion of COVID-19

Journal of Clinical Medicine ◽

10.3390/jcm10010004 ◽

2020 ◽

Vol 10 (1) ◽

pp. 4

Author(s):

Romain Vial ◽

Marion Gully ◽

Mickael Bobot ◽

Violaine Scarfoglière ◽

Philippe Brunet ◽

...

Keyword(s):

Predictive Value ◽

Eosinophil Count ◽

Characteristic Curve ◽

Final Analysis ◽

Hemodialysis Patients ◽

Chronic Hemodialysis ◽

Screening Methods ◽

Dialysis Session ◽

Diagnostic Study ◽

Rt Pcr

Background: Daily management to shield chronic dialysis patients from SARS-CoV-2 contamination makes patient care cumbersome. There are no screening methods to date and a molecular biology platform is essential to perform RT-PCR for SARS-CoV-2; however, accessibility remains poor. Our goal was to assess whether the tools routinely used to monitor our hemodialysis patients could represent reliable and quickly accessible diagnostic indicators to improve the management of our hemodialysis patients in this pandemic environment. Methods: In this prospective observational diagnostic study, we recruited patients from La Conception hospital. Patients were eligible for inclusion if suspected of SARS-CoV-2 infection when arriving at our center for a dialysis session between March 12th and April 24th 2020. They were included if both RT-PCR result for SARS-CoV-2 and cell blood count on the day that infection was suspected were available. We calculated the area under the curve (AUC) of the receiver operating characteristic curve. Results: 37 patients were included in the final analysis, of which 16 (43.2%) were COVID-19 positive. For the day of suspected COVID-19, total leukocytes were significantly lower in the COVID-19 positive group (4.1 vs. 7.4 G/L, p = 0.0072) and were characterized by lower neutrophils (2.7 vs. 5.1 G/L, p = 0.021) and eosinophils (0.01 vs. 0.15 G/L, p = 0.0003). Eosinophil count below 0.045 G/L identified SARS-CoV-2 infection with AUC of 0.9 [95% CI 0.81—1] (p < 0.0001), sensitivity of 82%, specificity of 86%, a positive predictive value of 82%, a negative predictive value of 86% and a likelihood ratio of 6.04. Conclusions: Eosinophil count enables rapid routine screening of symptomatic chronic hemodialysis patients suspected of being COVID-19 within a range of low or high probability.

Download Full-text

Soluble angiotensin-converting enzyme 2 as a prognostic biomarker for disease progression in patients infected with SARS-CoV-2

10.1101/2021.10.13.21264901 ◽

2021 ◽

Author(s):

Noelia Diaz Troyano ◽

Pablo Gabriel Medina ◽

Stephen Weber ◽

Martin Klammer ◽

Raquel Barquin-DelPino ◽

...

Keyword(s):

Angiotensin Converting Enzyme ◽

Disease Severity ◽

Predictive Value ◽

Characteristic Curve ◽

Converting Enzyme ◽

Rt Pcr ◽

Angiotensin Converting Enzyme 2 ◽

Reactive Protein ◽

Patient Groups ◽

Negative Controls

Background: There is a need for better prediction of disease severity in patients infected with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Soluble angiotensin-converting enzyme 2 (sACE2) arises from shedding of membrane ACE2 (mACE2) that is known to be a receptor for the spike protein of SARS-CoV-2; however, its value as a biomarker for disease severity is unknown. This study evaluated the predictive value of sACE2 in the context of other known biomarkers of inflammation and tissue damage (C-reactive protein [CRP], growth/differentiation factor-15 [GDF-15], interleukin-6 [IL-6], and soluble fms-like tyrosine kinase-1 [sFlt-1]) in patients with and without SARS-CoV-2 with different clinical outcomes. Methods: For univariate analyses, median differences between biomarker levels were calculated for the following patient groups classified according to clinical outcome: reverse transcription polymerase chain reaction (RT-PCR)-confirmed SARS-CoV-2 positive (Groups 1 – 4); RT-PCR-confirmed SARS-CoV-2 negative following previous SARS-CoV-2 infection (Groups 5 and 6); and RT-PCR-confirmed SARS-CoV-2 negative controls (Group 7). Results: Median levels of CRP, GDF-15, IL-6, and sFlt-1 were significantly higher in patients with SARS-CoV-2 who were admitted to hospital compared with patients who were discharged (all p<0.001), whereas levels of sACE2 were significantly lower (p<0.001). Receiver operating characteristic curve analysis of sACE2 provided cut-offs for the prediction of hospital admission of ≤0.05 ng/mL (positive predictive value: 89.1%) and ≥0.42 ng/mL (negative predictive value: 84.0%). Conclusion: These findings support further investigation of sACE2, either as a single biomarker or as part of a panel, to predict hospitalisation risk and disease severity in patients infected with SARS-CoV-2.

Download Full-text

Triage of Patients Suspected of COVID-19 in Chronic Hemodialysis: Eosinophil Count Differentiates Low and High Suspicion of COVID-19

10.21203/rs.3.rs-58052/v1 ◽

2020 ◽

Author(s):

Romain Vial ◽

Marion Gully ◽

Mickael Bobot ◽

Violaine Scarfoglière ◽

Philippe Brunet ◽

...

Keyword(s):

Predictive Value ◽

Eosinophil Count ◽

Characteristic Curve ◽

Final Analysis ◽

Hemodialysis Patients ◽

Chronic Hemodialysis ◽

Screening Methods ◽

Dialysis Session ◽

Diagnostic Study ◽

Rt Pcr

Abstract Background: Daily management to shield chronic dialysis patients from SARS-CoV-2 contamination makes patient care cumbersome. There are no screening methods to date and a molecular biology platform is essential to perform RT-PCR for SARS-CoV-2; however, accessibility remains poor. Our goal was to assess whether the tools routinely used to monitor our hemodialysis patients could represent reliable and quickly accessible diagnostic indicators to improve the management of our hemodialysis patients in this pandemic environment.Methods: In this prospective observational diagnostic study, we recruited patients from La Conception hospital. Patients were eligible for inclusion if suspected of SARS-CoV-2 infection when arriving at our center for a dialysis session between March 12th and April 24th 2020. They were included if both RT-PCR result for SARS-CoV-2 and cell blood count on the day that infection was suspected were available. We calculated the area under the curve (AUC) of the receiver operating characteristic curve.Results: 37 patients were included in the final analysis, of which 16 (43.2%) were COVID-19 positive. For the day of suspected COVID-19, total leukocytes were significantly lower in the COVID-19 positive group (4.1 vs 7.4 G/L, p=0.0072) and were characterized by lower neutrophils (2.7 vs 5.1 G/L, p=0.021) and eosinophils (0.01 vs 0.15 G/L, p=0.0003). Eosinophil count below 0.045 G/L identified SARS-CoV-2 infection with AUC of 0.9 [95% CI 0.81-1] (p<0.0001), sensitivity of 82%, specificity of 86%, a positive predictive value of 82%, a negative predictive value of 86% and a likelihood ratio of 6.04.Conclusions :Eosinophil count enables rapid routine screening of chronic hemodialysis patients suspected of being COVID-19 within a range of low or high probability.

Download Full-text

The Role of a Decision Rule in Symptomatic Pulmonary Embolism Patients with a Non-high Probability Ventilation-perfusion Scan

Thrombosis and Haemostasis ◽

10.1055/s-0038-1657630 ◽

1997 ◽

Vol 78 (02) ◽

pp. 794-798 ◽

Cited By ~ 2

Author(s):

Bowine C Michel ◽

Philomeen M M Kuijer ◽

Joseph McDonnell ◽

Edwin J R van Beek ◽

Frans F H Rutten ◽

...

Keyword(s):

Pulmonary Embolism ◽

Decision Rule ◽

Predictive Value ◽

Patient Population ◽

High Probability ◽

Characteristic Curve ◽

Stepwise Logistic Regression ◽

Pulmonary Angiogram ◽

Symptomatic Pulmonary Embolism ◽

Perfusion Scan

Summary Background: In order to improve the use of information contained in the medical history and physical examination in patients with suspected pulmonary embolism and a non-high probability ventilation-perfusion scan, we assessed whether a simple, quantitative decision rule could be derived for the diagnosis or exclusion of pulmonary embolism. Methods: In 140 consecutive symptomatic patients with a non- high probability ventilation-perfusion scan and an interpretable pulmonary angiogram, various clinical and lung scan items were collected prospectively and analyzed by multivariate stepwise logistic regression analysis to identify the most informative combination of items. Results: The prevalence of proven pulmonary embolism in the patient population was 27.1%. A decision rule containing the presence of wheezing, previous deep venous thrombosis, recently developed or worsened cough, body temperature above 37° C and multiple defects on the perfusion scan was constructed. For the rule the area under the Receiver Operating Characteristic curve was larger than that of the prior probability of pulmonary embolism as assessed by the physician at presentation (0.76 versus 0.59; p = 0.0097). At the cut-off point with the maximal positive predictive value 2% of the patients scored positive, at the cut-off point with the maximal negative predictive value pulmonary embolism could be excluded in 16% of the patients. Conclusions: We derived a simple decision rule containing 5 easily interpretable variables for the patient population specified. The optimal use of the rule appears to be in the exclusion of pulmonary embolism. Prospective validation of this rule is indicated to confirm its clinical utility.

Download Full-text

Machine Learning Prediction of SARS-CoV-2 Polymerase Chain Reaction Results with Routine Blood Tests

Laboratory Medicine ◽

10.1093/labmed/lmaa111 ◽

2020 ◽

Author(s):

Thomas Tschoellitsch ◽

Martin Dünser ◽

Carl Böck ◽

Karin Schwarzbauer ◽

Jens Meier

Keyword(s):

Machine Learning ◽

Polymerase Chain Reaction ◽

Characteristic Curve ◽

Cohort Analysis ◽

Rt Pcr ◽

Chain Reaction ◽

Blood Tests ◽

Routine Blood ◽

Machine Learning Model ◽

Polymerase Chain

Abstract Objective The diagnosis of COVID-19 is based on the detection of SARS-CoV-2 in respiratory secretions, blood, or stool. Currently, reverse transcription polymerase chain reaction (RT-PCR) is the most commonly used method to test for SARS-CoV-2. Methods In this retrospective cohort analysis, we evaluated whether machine learning could exclude SARS-CoV-2 infection using routinely available laboratory values. A Random Forests algorithm with 1353 unique features was trained to predict the RT-PCR results. Results Out of 12,848 patients undergoing SARS-CoV-2 testing, routine blood tests were simultaneously performed in 1528 patients. The machine learning model could predict SARS-CoV-2 test results with an accuracy of 86% and an area under the receiver operating characteristic curve of 0.90. Conclusion Machine learning methods can reliably predict a negative SARS-CoV-2 RT-PCR test result using standard blood tests.

Download Full-text

Machine learning approach for differentiating cytomegalovirus esophagitis from herpes simplex virus esophagitis

Scientific Reports ◽

10.1038/s41598-020-78556-z ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Jung Su Lee ◽

Jihye Yun ◽

Sungwon Ham ◽

Hyunjung Park ◽

Hyunsu Lee ◽

...

Keyword(s):

Machine Learning ◽

Differential Diagnosis ◽

Herpes Simplex Virus ◽

Herpes Simplex ◽

Predictive Value ◽

Characteristic Curve ◽

Previous History ◽

Clinical Factor ◽

Endoscopic Images ◽

Simplex Virus

AbstractThe endoscopic features between herpes simplex virus (HSV) and cytomegalovirus (CMV) esophagitis overlap significantly, and hence the differential diagnosis between HSV and CMV esophagitis is sometimes difficult. Therefore, we developed a machine-learning-based classifier to discriminate between CMV and HSV esophagitis. We analyzed 87 patients with HSV esophagitis and 63 patients with CMV esophagitis and developed a machine-learning-based artificial intelligence (AI) system using a total of 666 endoscopic images with HSV esophagitis and 416 endoscopic images with CMV esophagitis. In the five repeated five-fold cross-validations based on the hue–saturation–brightness color model, logistic regression with a least absolute shrinkage and selection operation showed the best performance (sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and area under the receiver operating characteristic curve: 100%, 100%, 100%, 100%, 100%, and 1.0, respectively). Previous history of transplantation was included in classifiers as a clinical factor; the lower the performance of these classifiers, the greater the effect of including this clinical factor. Our machine-learning-based AI system for differential diagnosis between HSV and CMV esophagitis showed high accuracy, which could help clinicians with diagnoses.

Download Full-text

A meta-analysis of accuracy and sensitivity of chest CT and RT-PCR in COVID-19 diagnosis

Scientific Reports ◽

10.1038/s41598-020-80061-2 ◽

2020 ◽

Vol 10 (1) ◽

Author(s):

Fatemeh Khatami ◽

Mohammad Saatchi ◽

Seyed Saeed Tamehri Zadeh ◽

Zahra Sadat Aghamir ◽

Alireza Namazi Shabestari ◽

...

Keyword(s):

Polymerase Chain Reaction ◽

Reverse Transcriptase ◽

Ct Scan ◽

Predictive Value ◽

Meta Analysis ◽

Chest Ct ◽

Rt Pcr ◽

Chain Reaction ◽

Polymerase Chain ◽

Chest Ct Scan

AbstractNowadays there is an ongoing acute respiratory outbreak caused by the novel highly contagious coronavirus (COVID-19). The diagnostic protocol is based on quantitative reverse-transcription polymerase chain reaction (RT-PCR) and chests CT scan, with uncertain accuracy. This meta-analysis study determines the diagnostic value of an initial chest CT scan in patients with COVID-19 infection in comparison with RT-PCR. Three main databases; PubMed (MEDLINE), Scopus, and EMBASE were systematically searched for all published literature from January 1st, 2019, to the 21st May 2020 with the keywords "COVID19 virus", "2019 novel coronavirus", "Wuhan coronavirus", "2019-nCoV", "X-Ray Computed Tomography", "Polymerase Chain Reaction", "Reverse Transcriptase PCR", and "PCR Reverse Transcriptase". All relevant case-series, cross-sectional, and cohort studies were selected. Data extraction and analysis were performed using STATA v.14.0SE (College Station, TX, USA) and RevMan 5. Among 1022 articles, 60 studies were eligible for totalizing 5744 patients. The overall sensitivity, specificity, positive predictive value, and negative predictive value of chest CT scan compared to RT-PCR were 87% (95% CI 85–90%), 46% (95% CI 29–63%), 69% (95% CI 56–72%), and 89% (95% CI 82–96%), respectively. It is important to rely on the repeated RT-PCR three times to give 99% accuracy, especially in negative samples. Regarding the overall diagnostic sensitivity of 87% for chest CT, the RT-PCR testing is essential and should be repeated to escape misdiagnosis.

Download Full-text

CEUS detection of biliary ischaemia during the first 4 weeks after liver transplantation predicts non-anastomotic biliary stricture

Clinical Hemorheology and Microcirculation ◽

10.3233/ch-211097 ◽

2021 ◽

pp. 1-12

Author(s):

Bo-wen Zheng ◽

Shu-hong Yi ◽

Tao Wu ◽

Mei Liao ◽

Ying-cai Zhang ◽

...

Keyword(s):

Liver Transplantation ◽

Predictive Value ◽

Biliary Stricture ◽

Characteristic Curve ◽

Validation Group ◽

Gamma Glutamyl Transpeptidase ◽

Full Cohort ◽

Non Invasive ◽

Invasive Method ◽

Nas Score

BACKGROUND: Biliary ischaemia is an important factor in the pathogenesis of non-anastomotic biliary stricture (NAS) after liver transplantation (LT). Contrast-enhanced ultrasound (CEUS) can be used to detect biliary ischaemia, but no study has examined the utility of CEUS in predicting NAS. OBJECTIVE: To evaluate whether repeated CEUS as a non-invasive method of biliary ischaemia can identify NAS. METHODS: Consecutive LT patients who underwent CEUS examinations at 1–4 weeks after LT from September 2012 to December 2015 at our institution were included. The CEUS images and clinical data were analysed. RESULTS: Among 116 eligible LT patients, 39 (33.6%) were diagnosed with NAS within 1 year after LT. The patients with NAS had a significantly higher CEUS score at weeks 2–4 (all P < 0.05) and a higher slope of CEUS score progression (0.480 vs –0.044, P < 0.001). The accuracy of CEUS in identifying NAS improved over time after LT, reaching its maximum at week 4, with a sensitivity of 66.7%, a specificity of 87.9%, a positive predictive value (PPV) of 75.9%, a negative predictive value (NPV) of 82.3%, and an accuracy of 80.2%in the full cohort when a CEUS score≥3 was used as the cut-off. Multivariate analysis identified gamma-glutamyl transpeptidase (GGT), alanine transaminase (ALT) and the CEUS score at week 4 as independent predictors of NAS. In the task of identifying NAS, an NAS score combining the above 3 variables at week 4 showed areas under the receiver operating characteristic curve of 0.88 (95%CI, 0.78–0.99) in the estimation group (n = 60) and 0.82 (95%CI, 0.69–0.96) in the validation group (n = 56). An NAS score cut-off of 0.396 identified 87.2%of NAS cases in the estimation group, with a PPV of 93.3%; and 75.0%of NAS cases in the validation group, with a PPV of 58.8%. CONCLUSIONS: CEUS examination during the first 4 weeks is useful in assessing the risk of NAS within 1 year after LT. In particular, an NAS score combining the CEUS score, GGT level, and ALT level at week 4 can be used to accurately predict the risk of NAS in LT patients.

Download Full-text

Aggregation of cohorts for histopathological diagnosis with deep morphological analysis

Scientific Reports ◽

10.1038/s41598-021-82642-1 ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Jeonghyuk Park ◽

Yul Ri Chung ◽

Seo Taek Kong ◽

Yeong Won Kim ◽

Hyunho Park ◽

...

Keyword(s):

Cancer Detection ◽

Morphological Analysis ◽

Deep Neural Networks ◽

Large Datasets ◽

Histopathological Diagnosis ◽

Single Model ◽

Trade Off ◽

Detection Model ◽

Optimal Behavior ◽

The Cost

AbstractThere have been substantial efforts in using deep learning (DL) to diagnose cancer from digital images of pathology slides. Existing algorithms typically operate by training deep neural networks either specialized in specific cohorts or an aggregate of all cohorts when there are only a few images available for the target cohort. A trade-off between decreasing the number of models and their cancer detection performance was evident in our experiments with The Cancer Genomic Atlas dataset, with the former approach achieving higher performance at the cost of having to acquire large datasets from the cohort of interest. Constructing annotated datasets for individual cohorts is extremely time-consuming, with the acquisition cost of such datasets growing linearly with the number of cohorts. Another issue associated with developing cohort-specific models is the difficulty of maintenance: all cohort-specific models may need to be adjusted when a new DL algorithm is to be used, where training even a single model may require a non-negligible amount of computation, or when more data is added to some cohorts. In resolving the sub-optimal behavior of a universal cancer detection model trained on an aggregate of cohorts, we investigated how cohorts can be grouped to augment a dataset without increasing the number of models linearly with the number of cohorts. This study introduces several metrics which measure the morphological similarities between cohort pairs and demonstrates how the metrics can be used to control the trade-off between performance and the number of models.

Download Full-text