When Does Bone Suppression and Lung Field Segmentation Improve Chest X-Ray Disease Classification?
Standard
When Does Bone Suppression and Lung Field Segmentation Improve Chest X-Ray Disease Classification? / Baltruschat, Ivo Matteo; Steinmeister, Leonhard; Ittrich, Harald; Adam, Gerhard; Nickisch, Hannes; Saalbach, Axel; von Berg, J; Grass, Michael; Knopp, Tobias.
IEEE: 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019) . 2019. S. 1362-1366.Publikationen: SCORING: Beitrag in Buch/Sammelwerk › Konferenzbeitrag - Aufsatz in Konferenzband › Forschung › Begutachtung
Harvard
APA
Vancouver
Bibtex
}
RIS
TY - CHAP
T1 - When Does Bone Suppression and Lung Field Segmentation Improve Chest X-Ray Disease Classification?
AU - Baltruschat, Ivo Matteo
AU - Steinmeister, Leonhard
AU - Ittrich, Harald
AU - Adam, Gerhard
AU - Nickisch, Hannes
AU - Saalbach, Axel
AU - von Berg, J
AU - Grass, Michael
AU - Knopp, Tobias
PY - 2019
Y1 - 2019
N2 - Chest radiography is the most common clinical examination type. To improve the quality of patient care and to reduce workload, methods for automatic pathology classificationhave been developed. In this contribution we investigate theusefulness of two advanced image pre-processing techniques,initially developed for image reading by radiologists, for theperformance of Deep Learning methods. First, we use bonesuppression, an algorithm to artificially remove the rib cage. Secondly, we employ an automatic lung field detection to crop the image to the lung area. Furthermore, we consider the combination of both in the context of an ensemble approach. In a five-times re-sampling scheme, we use Receiver Op-erating Characteristic (ROC) statistics to evaluate the effect ofthe pre-processing approaches. Using a Convolutional NeuralNetwork (CNN), optimized for X-ray analysis, we achieve a good performance with respect to all pathologies on average. Superior results are obtained for selected pathologies when using pre-processing, i.e. for mass the area under the ROC curve increased by 9.95%. The ensemble with pre-processed trained models yields the best overall results.
AB - Chest radiography is the most common clinical examination type. To improve the quality of patient care and to reduce workload, methods for automatic pathology classificationhave been developed. In this contribution we investigate theusefulness of two advanced image pre-processing techniques,initially developed for image reading by radiologists, for theperformance of Deep Learning methods. First, we use bonesuppression, an algorithm to artificially remove the rib cage. Secondly, we employ an automatic lung field detection to crop the image to the lung area. Furthermore, we consider the combination of both in the context of an ensemble approach. In a five-times re-sampling scheme, we use Receiver Op-erating Characteristic (ROC) statistics to evaluate the effect ofthe pre-processing approaches. Using a Convolutional NeuralNetwork (CNN), optimized for X-ray analysis, we achieve a good performance with respect to all pathologies on average. Superior results are obtained for selected pathologies when using pre-processing, i.e. for mass the area under the ROC curve increased by 9.95%. The ensemble with pre-processed trained models yields the best overall results.
M3 - Conference contribution - Article for conference
SP - 1362
EP - 1366
BT - IEEE
ER -