Boosting CNN performance for lung texture classification using connected filtering

Sebastian Tarando (1, 2, 3) , Catalin Fetita (4, 1, 2, 3) , Young-Wouk Kim (5) , Hyoun Cho (6) , Pierre Yves Brillet (5, 7)

1 TSP - ARTEMIS - Département Advanced Research And Techniques For Multidimensional Imaging Systems
2 SAMOVAR - Services répartis, Architectures, MOdélisation, Validation, Administration des Réseaux
3 CNRS - Centre National de la Recherche Scientifique
4 ARMEDIA-SAMOVAR - ARMEDIA
5 Service de radiologie [Avicenne]
6 CHU Pitié-Salpêtrière [AP-HP]
7 UP13 - Université Paris 13

Infiltrative lung diseases describe a large group of irreversible lung disorders requiring regular follow-up with CT imaging. Quantifying the evolution of the patient status imposes the development of automated classification tools for lung texture. This paper presents an original image pre-processing framework based on locally connected filtering applied in multiresolution, which helps improving the learning process and boost the performance of CNN for lung texture classification. By removing the dense vascular network from images used by the CNN for lung classification, locally connected filters provide a better discrimination between different lung patterns and help regularizing the classification output. The approach was tested in a preliminary evaluation on a 10 patient database of various lung pathologies, showing an increase of 10% in true positive rate (on average for all the cases) with respect to the state of the art cascade of CNNs for this task

Domaines: Traitement du signal et de l'image [eess.SP]; Pneumologie et système respiratoire

Sebastian Tarando, Catalin Fetita, Young-Wouk Kim, Hyoun Cho, Pierre Yves Brillet. Boosting CNN performance for lung texture classification using connected filtering. Medical Imaging 2018: Computer-Aided Diagnosis, Feb 2018, Houston, United States. pp.1057505 -, ⟨10.1117/12.2293093⟩. ⟨hal-01995970⟩