Title: Effsemble: faster, smaller and more accurate ensemble networks for thoracic disease classification

Authors: Arren Matthew C. Antioquia

Addresses: Center for Computational Imaging and Visual Innovations, De La Salle University, Malate, Manila, Philippines

Abstract: Convolutional Neural Networks (CNNs) are being adapted to various computer-aided diagnosis applications, including recognising thoracic diseases. To improve classification performance, recent solutions alter the structure of existing networks or require additional prior information for training. Other approaches demand massive computational requirement to increase classification accuracy. In this paper, we propose a family of efficient ensemble networks called Effsemble to accurately recognise thoracic diseases without additional layers, extra input data, or large computational overhead. Our proposed approach achieves the highest average AUROC score of 80.04% in the ChestX-ray14 data set. Moreover, our Effsemble-2 outperforms other state-of-the-art methods, while decreasing the parameter count by 31.7M and increasing inference speed by 2.75x compared to the previous best ensemble.

Keywords: thoracic disease classification; image classification; convolutional neural networks; ensemble learning; deep learning.

DOI: 10.1504/IJCAT.2023.132406

International Journal of Computer Applications in Technology, 2023 Vol.71 No.4, pp.332 - 339

Received: 19 May 2022
Received in revised form: 13 Sep 2022
Accepted: 15 Sep 2022
Published online: 19 Jul 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article