An ensemble of fine-tuned fully convolutional neural networks for pleural effusion cell nuclei segmentation. (January 2020)

• Record Type: Journal Article
• Authors: Baykal Kablan, Elif; Dogan, Hulya; Ercin, Mustafa Emre; Ersoz, Safak; Ekinci, Murat
• Is Part Of: Computers & electrical engineering. Volume 81(2020)
• Publication Date: 2020-01
• Subjects: Fully convolutional neural networks -- Nuclei segmentation -- Ensemble network -- Pleural effusion -- Cytology; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2019.106533
• Languages: English
• ISSNs: 0045-7906

Abstract

Abstract: Pleural effusion (PE) is a problem commonly encountered in patients with malignant neoplasms, especially carcinomas of the breast and lung. Nuclei segmentation is a prerequisite step of a computer-aided diagnosis (CAD) system for pleural cancer. We propose an ensemble of fully convolutional neural networks (FCNN) for PE cell nuclei segmentation. A two-stage method, fine-tuning of deep FCNN and an ensemble network, is designed. It involves, first, fine-tuning the well-known deep learning-based segmentation networks such as fully convolutional network (FCN), SegNet, and U-Net, and then employing an ensemble network that uses the features from fine-tuned networks to improve the efficiency. The proposed ensemble method achieves a Jaccard index of 90.82%, a sensitivity of 97.15%, and a specificity of 99.80% on the new dataset consisting of 120 PE cytology images. These experimental results have not only contributed to the pleural cancer diagnosis but also demonstrated the effectiveness of an ensemble of FCNN.