Research on identification methods for the Coronavirus Disease 2019 (COVID-19) has increased in the last years and the need for automated detection methods has surged as well. Computed Tomography scan images have demonstrated to contain useful and sufficient information to detect COVID-19 by using machine learning and computational intelligence techniques. However, in order to expand their adoption in medical clinics, COVID-19 detection approaches need to drive the experts in the overall comprehension of the classification, to check the validity and meaningfulness of the prediction results. Herein, we propose a deep learning approach based on an ensemble of convolutional neural networks with the aim of detecting, very accurately and in an explainable way, COVID-19 patients by leveraging CT scan images. We also take advantage of transfer learning and apply the aforementioned deep ensemble to a large publicly available dataset, by clustering the images per lung lobe. Our results show good classification performance, good generalization potentials, as well as quite interpretable outcomes.
Explainable Deep Ensemble to Diagnose COVID-19 from CT Scans / Aversano, Lerina; Bernardi, Mario Luca; Cimitile, Marta; Pecori, Riccardo; Verdone, Chiara. - 14069:(2023), pp. 642-654. (Intervento presentato al convegno 13th Conference of the European Society for Fuzzy Logic and Technology, EUSFLAT 2023, and 12th International Summer School on Aggregation Operators tenutosi a Palma de Maiorca, Spain nel September 2023) [10.1007/978-3-031-39965-7_53].
Explainable Deep Ensemble to Diagnose COVID-19 from CT Scans
Pecori, Riccardo
;
2023-01-01
Abstract
Research on identification methods for the Coronavirus Disease 2019 (COVID-19) has increased in the last years and the need for automated detection methods has surged as well. Computed Tomography scan images have demonstrated to contain useful and sufficient information to detect COVID-19 by using machine learning and computational intelligence techniques. However, in order to expand their adoption in medical clinics, COVID-19 detection approaches need to drive the experts in the overall comprehension of the classification, to check the validity and meaningfulness of the prediction results. Herein, we propose a deep learning approach based on an ensemble of convolutional neural networks with the aim of detecting, very accurately and in an explainable way, COVID-19 patients by leveraging CT scan images. We also take advantage of transfer learning and apply the aforementioned deep ensemble to a large publicly available dataset, by clustering the images per lung lobe. Our results show good classification performance, good generalization potentials, as well as quite interpretable outcomes.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
