Chest radiography is the most common clinical examination type. To improve the quality of patient care and to reduce workload, researchers started developing methods for automatic pathology classification. In this contribution we investigate the effect of advanced image processing techniques, initially developed to support radiologists (i.e. bone suppression and lung segmentation), in the context of an automatic approach for chest X-ray analysis (using Deep Learning).

In this study, we employ bone suppression [3], an algorithm to artificially remove the rib cage in chest X-ray images, and automatic lung field detection to crop images to the lung area, in order to study their usefulness in the context of Deep Learning. Furthermore, we consider the combination of both. For Convolutional Neural Network (CNN) training and evaluation, DICOM images from the Indiana dataset (Open-I)[1] were examined by two expert radiologists and annotated with respect to eight different pathologies: pleural effusion, infiltrate, congestion, atelectasis,...

In a five-fold cross-validation, we use ROC statistics to evaluate the effect of the preprocessing approaches. While the trained models show in general a good performance with respect to all categories, the combination of bone suppression and lung field detection improves the average area under the ROC curve slightly from 89.1±1.3 to 90.6±1.2. In contrast, for selected pathologies, a substantial improvement can be reported (i.e. “mass”: 76.4 ± 1.6 vs. 84.0 ± 1.1).

Advanced image pre-processing improves the overall CNN performance. Furthermore, for pathologies with small aspects, such as foreign objects and masses, Deep Learning models can benefit from the increased effective spatial resolution, resulting in a significantly improved outcome.

PD Dr. habil. Michael Graß Principal Scientist Philips Research Philips GmbH Innovative Technologies Research Laboratories Hamburg Röntgenstrasse 24-26 22335 Hamburg / Germany

[1] Dina Demner-Fushman, Marc D. Kohli, Marc B.Rosenman, Sonya E. Shooshan, Laritza Rodriguez, Sameer Antani, George R. Thoma, and Clement J. McDonald, “Preparing a collection of radiology examinations for distribution and retrieval,” Journal of the American Medical Informatics Association, 2015 [2] Xiaosong Wang, Yifan Peng, Le Lu, Zhiyong Lu, Mohammadhadi Bagheri, and Ronald M. Summers, “ChestX-Ray8: Hospital-scale chest X-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases,” in Conference on Computer Vision and Pattern Recognition, 2017 [3] Jens von Berg, Claire...