L’équipe ARMEDIA vous convie à un séminaire dans la thématique « IDIA et santé »
le jeudi 25 juin à 14h00.
L’exposé sera présenté en visio, et, si les conditions le permettent, simultanément en face-à-face sur le site de Telecom SudParis, salle H218 (nombre limité de places pour respecter les conditions sanitaires ; port de masque imposé…).

Le lien visio pour le séminaire est le suivant :

https://webconf.imt.fr/frontend/cat-clh-6al

Les détails de l’exposé présenté par Catalin FETITA figurent ci-dessous.

Title: Comparison of CNN architectures and training strategies for quantitative analysis of idiopathic interstitial pneumonia. Projection to COVID follow-up strategies.

Abstract. Fibrosing idiopathic interstitial pneumonia (IIP) is a subclass of interstitial lung diseases manifesting as progressive worsening of lung function. Such degradation is a continuous and irreversible process which requires quantitative follow-up of patients to assess the pathology occurrence and extent in the lung. The development of automated CAD tools for such purpose is oriented today towards machine learning approaches and in particular convolutional neural networks. The difficulty remains in the choice of the network architecture that best fit to the problem, in straight relationship with available databases for training. In this work we investigate two CNN architectures and different training strategies in the context of a limited database, with high class imbalance and subjective and partial annotations. We show that increased performances are achieved using an end-to-end architecture versus patch-based, but also that naive implementation in the former case should be avoided. The proposed solution is able to leverage global information in the scan and shows a high improvement in the F1 scores of the predicted classes and visual results of predictions in better accordance with the radiologist expectations. The end of presentation will highlight potential applications of the developed solution to COVID-19 patients follow-up, together with additional image biomarkers related to vascular remodeling and lung compliance.