Content-Based Image Retrieval of Chest CT with Convolutional Neural Network for Diffuse Interstitial Lung Disease: Performance Assessment in Three Major Idiopathic Interstitial Pneumonias
Authors
Hye Jeon Hwang, MD, PhD, Joon Beom Seo, MD, PhD, Sang Min Lee, MD, PhD, Eun Young Kim, MD, PhD, Beomhee Park, MS, Hyun-Jin Bae, PhD, Namkug Kim, PhD
In this study, the performance of a content-based image retrieval (CBIR) system for diffuse interstitial lung disease (DILD) was evaluated. The CBIR system utilized a convolutional neural network to automatically quantify and classify six image patterns of DILD from chest CT scans. A total of 246 pairs of initial and follow-up chest CT scans from patients with usual interstitial pneumonia, nonspecific interstitial pneumonia, and cryptogenic organic pneumonia were used to assess the CBIR system's performance. Sixty cases were selected as queries, and the CBIR system retrieved five similar CT scans from the database. The results showed that the CBIR system had a high rate of retrieving the same pairs of query CT scans in the top 1-5 retrievals and could retrieve similar CT scans more accurately in UIP compared to NSIP and COP. Radiologists also rated a high percentage of retrieved CT scans as having a high degree of similarity to the query CT scans. The study demonstrates the potential utility of the CBIR system for improving the diagnosis and management of DILD, and highlights the potential for commercialization of AVIEW software.