TYPE: Late Breaking Abstract
TOPIC: Lung Pathology
PURPOSE: Comparison of three different AI-based software for quantification of lung parenchyma affected on Covid-19 parients
METHODS: This double-center study includes 120 COVID-19 patients (60 from each center) with positive reverse-transcription polymerase chain reaction (RT-PCR) who underwent a chest CT scan from November 2020 to February 2021. CT scans were analyzed retrospectively and independently in each center. CT scans were examined by two different Radiologists for each center, providing the qualitative score of lung involvement, whereas the quantitative analysis was performed by one trained Radiographer for each center using three different software: 3DSlicer, CT Lung Density Analysis and CT Pulmo 3D
RESULTS: The agreement between Radiologists for visual estimation of pneumonia at CT was good (ICC 0.79, 95% CI 0,73-0,84). 3DSlicer had an over-esteem of the measures assessed, however ICC index returned a value of 0.92 (CI 0,90-0,94), indicating an excellent reliability within the three software employed. The ICC was performed between each single software and the median of the visual score provided by the Radiologists. The best agreement was between 3DSlicer and the median of the visual score (0.75 with a CI 0,67 - 0,82 and with a median value of 22% of disease extension for the software and 25% the visual values).
CONCLUSIONS: Visual and software results correlate positively and 3D slicer is the one that predominantly fitted the Radiologists scores. Additionally, AI-based lung segmentation’s inter-software output correlation is even stronger and they all were found to overestimate the amount of lung parenchyma involved in comparison of visual score.
CLINICAL IMPLICATIONS: Clinical management
DISCLOSURE: No significant relationships.
KEYWORD: Artificial Intelligence