eduzhai > Applied Sciences > Engineering >

Comparative study of deep learning methods for the automatic segmentation of lung lesion and lesion type in CT scans of COVID-19 patients

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 21 pages

Abstract: Recent research on COVID-19 suggests that CT imaging provides usefulinformation to assess disease progression and assist diagnosis, in addition tohelp understanding the disease. There is an increasing number of studies thatpropose to use deep learning to provide fast and accurate quantification ofCOVID-19 using chest CT scans. The main tasks of interest are the automaticsegmentation of lung and lung lesions in chest CT scans of confirmed orsuspected COVID-19 patients. In this study, we compare twelve deep learningalgorithms using a multi-center dataset, including both open-source andin-house developed algorithms. Results show that ensembling different methodscan boost the overall test set performance for lung segmentation, binary lesionsegmentation and multiclass lesion segmentation, resulting in mean Dice scoresof 0.982, 0.724 and 0.469, respectively. The resulting binary lesions weresegmented with a mean absolute volume error of 91.3 ml. In general, the task ofdistinguishing different lesion types was more difficult, with a mean absolutevolume difference of 152 ml and mean Dice scores of 0.369 and 0.523 forconsolidation and ground glass opacity, respectively. All methods performbinary lesion segmentation with an average volume error that is better thanvisual assessment by human raters, suggesting these methods are mature enoughfor a large-scale evaluation for use in clinical practice.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×