Determination of disease severity in COVID-19 patients using deep learning in chest X-ray images

Blain, Maxime; Kassin, Michael; Varble, Nicole; Wang, Xiaosong; Xu, Ziyue; Xu, Daguang; Carrafiello, Gianpaolo; Vespro, Valentina; Stellato, Elvira; Meglio, Letizia Di; Suh, Robert D.; Walker, Stephanie A.; Xu, Sheng; Sanford, Thomas; Türkbey, Evrim; Harmon, Stephanie; Türkbey, Barış; Wood, Bradford J.

doi:10.5152/dir.2020.20205

Cited by 52 publications

(40 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…However, determining the extent of Covid-19 pneumonia on X-ray is limited since the sensitivity for ground-glass opacity is imperfect. Recently published works studied the use of artificial intelligence as a tool to increase X-ray diagnostic and prognostic power, which is a great support especially in places where CT is not readily available [25,26].…”

Section: Discussionmentioning

confidence: 99%

Determining extent of COVID-19 pneumonia on CT based on biological variables

Tordjman

Mekki

Mali

et al. 2020

Respiratory Medicine

View full text Add to dashboard Cite

Covid-19 pneumonia CT extent correlates well with outcome including mortality. However, CT is not widely available in many countries. This study aimed to explore the relationship between Covid-19 pneumonia CT extent and blood tests variations. The objective was to determine for the biological variables correlating with disease severity the cutoff values showing the best performance to predict the parenchymal extent of the pneumonia. Methods: Bivariate correlations were calculated between biological variables and grade of disease extent on CT. Receiving Operating Characteristic curve analysis determined the best cutoffs for the strongest correlated biological variables. The performance of these variables to predict mild (<10%) or severe pneumonia (>50% of parenchyma involved) was evaluated. Results: Correlations between biological variables and disease extent was evaluated in 168 patients included in this study. LDH, lymphocyte count and CRP showed the strongest correlations (with 0.67, − 0.41 and 0.52 correlation coefficient, respectively). Patients were split into a training and a validation cohort according to their centers. If one variable was above/below the following cutoffs , LDH>380, CRP>80 or lymphocyte count <0.8G/L, severe pneumonia extent on CT was detected with 100% sensitivity. Values above/below all three thresholds were denoted in 73% of patients with severe pneumonia extent. The combination of LDH<220 and CRP<22 was associated with mild pneumonia extent (<10%) with specificity of 100%. Discussion: LDH showed the strongest correlation with the extent of Covid-19 pneumonia on CT. Combined with CRP±lymphocyte count, it helps predicting parenchymal extent of the pneumonia when CT scan is not available.

show abstract

Section: Discussionmentioning

confidence: 99%

Determining extent of COVID-19 pneumonia on CT based on biological variables

Tordjman

Mekki

Mali

et al. 2020

Respiratory Medicine

View full text Add to dashboard Cite

show abstract

“…The maturity of the papers was assessed by co-authors who have developed or worked with DL algorithms (see Figure 5(middle)). Only 13 highly mature studies were identified 15,16,17,18,19,20,21,22,23,24,25,26,27 . The list of papers and their task and modality appear in Table 1.…”

Section: Resultsmentioning

confidence: 99%

“…The CT findings correlated with several worse symptoms, including a respiratory rate greater than 30 breaths per minute, and oxygen saturation of 93% or less in a resting state among other phenotypes 39 . In clinical practice, often progress assessments as well as patient management is performed based on CXR and not chest CT. AI that provides assessment of severity could be useful if it was quantifiable and accurate and 4 publications were found mature in performing this task 23,24,25,26 . One of them utilized a dataset of multiple CT scans per patients and introduced a “CT scan simulator” that modelled the temporal evolution of the CT through disease progression and was evaluated on multi-national and multi-machine data 25 .…”

Section: Resultsmentioning

confidence: 99%

“…Notably, despite the overall overhead of CXR compared to CT in the analysed publications, only 4 publications in the meta-analysis focused on severity assessment based on CXR (20 on CT, 1 on LUS). One of them was found mature – and trained DL models on lung segmentation and opacity detection of 48 COVID-19 patients and achieved an agreement measure (Cohen’s kappa) of 0.51 for alveolar opacities and 0.71 for interstitial opacities 26 . In the only identified publication with multimodal imaging data for one patient cohort 14 , manual airspace disease segmentation of CTs in 86 COVID-19 patients was used as ground truth to train an AI on volumetric quantification from CXR.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

On the Role of Artificial Intelligence in Medical Imaging of COVID-19

Born

Beymer

Rajan

et al. 2020

Preprint

View full text Add to dashboard Cite

During the COVID-19 pandemic, lung imaging takes a key role in addressing the magnified need of speed, cost, ubiquity and precision in medical care. The rise of artificial intelligence induced a quantum leap in medical imaging: AI has now proven equipollent to healthcare professionals in several diseases and the potential to save time, cost and increase coverage. But AI-accelerated medical imaging must still fully demonstrate its ability in remediating diseases such as COVID-19. We identify key use cases of lung imaging for COVID-19, comparing CT, X-Ray and ultrasound imaging from clinical and AI perspectives. We perform a systematic, manual survey of 197 related publications that reveals a disparity in the focus of the AI and clinical communities, caused by data availability and the lack of collaboration, and in modality trends, driven by ubiquity. Last, challenges in AI-acceleration and ways to remediate them are discussed and future research goals are identified.

show abstract

“…In addition to the PXS score model used for the basis of the AI-guided severity grading in this study, multiple deep learning-based models have been published that also automatically assess COVID-19 lung disease severity from CXR data ( 11 , 12 , 13 , 14 , 15 , 16 , 17 ). The next logical step in the evolution of such an AI system would be to report the AI-based severity score by itself to the frontline clinicians, potentially bypassing the diagnostic radiologist.…”

Section: Discussionmentioning

confidence: 99%

Multi-Radiologist User Study for Artificial Intelligence-Guided Grading of COVID-19 Lung Disease Severity on Chest Radiographs

Little

Alkasab

et al. 2021

Academic Radiology

View full text Add to dashboard Cite

Rationale and Objectives: Radiographic findings of COVID-19 pneumonia can be used for patient risk stratification; however, radiologist reporting of disease severity is inconsistent on chest radiographs (CXRs). We aimed to see if an artificial intelligence (AI) system could help improve radiologist interrater agreement. Materials and Methods:We performed a retrospective multi-radiologist user study to evaluate the impact of an AI system, the PXS score model, on the grading of categorical COVID-19 lung disease severity on 154 chest radiographs into four ordinal grades (normal/minimal, mild, moderate, and severe). Four radiologists (two thoracic and two emergency radiologists) independently interpreted 154 CXRs from 154 unique patients with COVID-19 hospitalized at a large academic center, before and after using the AI system (median washout time interval was 16 days). Three different thoracic radiologists assessed the same 154 CXRs using an updated version of the AI system trained on more imaging data. Radiologist interrater agreement was evaluated using Cohen and Fleiss kappa where appropriate. The lung disease severity categories were associated with clinical outcomes using a previously published outcomes dataset using Fisher's exact test and Chi-square test for trend.Results: Use of the AI system improved radiologist interrater agreement (Fleiss k = 0.40 to 0.66, before and after use of the system). The Fleiss k for three radiologists using the updated AI system was 0.74. Severity categories were significantly associated with subsequent intubation or death within 3 days. Conclusion:An AI system used at the time of CXR study interpretation can improve the interrater agreement of radiologists.

show abstract

Determination of disease severity in COVID-19 patients using deep learning in chest X-ray images

Cited by 52 publications

References 35 publications

Determining extent of COVID-19 pneumonia on CT based on biological variables

Determining extent of COVID-19 pneumonia on CT based on biological variables

On the Role of Artificial Intelligence in Medical Imaging of COVID-19

Multi-Radiologist User Study for Artificial Intelligence-Guided Grading of COVID-19 Lung Disease Severity on Chest Radiographs

Contact Info

Product

Resources

About