Detecting Multi-Class Artifacts in Endoscopic Images using YOLOv3
1N.Kirthika*, Research scholar, Department of Electronics and Communication Engineering, School of Engineering, Avinashilingam Institute for Home Science and Higher Education for Women, Coimbatore, India.
2Dr.B.Sargunam, Associate Professor & Head, Department of Electronics and Communication Engineering, School of Engineering, Avinashilingam Institute for Home Science and Higher Education for Women, Coimbatore, India.
Manuscript received on May 18, 2020. | Revised Manuscript received on May 27, 2020. | Manuscript published on May 30, 2020. | PP: 2578-2583 | Volume-9 Issue-1, May 2020. | Retrieval Number: A3050059120/2020©BEIESP | DOI: 10.35940/ijrte.A3050.059120
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Endoscopy is a regular, clinical non-surgical procedure to examine hollow organs like esophagus, stomach, intestine, etc., Nowadays, due to rapid technological development and miniaturization of hardware endoscopy can be performed in the respiratory tract, urinary tract, female reproductive tract and joints apart from gastrointestinal (GI) tract. The need for GI endoscopy is to examine, and diagnose ailments like cancers, polyps, and assist cauterizing bleeding vessels. The organ nature creates lots of artifacts in the imaged tissue such as saturation, specularity, blur, contrast, bubbles, and debris which causes significant challenges in quantitative analysis. Thus, combining state-of-the-art deep learning object detection algorithms like YOLOv3 with endoscopy leads to efficient and accurate localization and categorization of different imaging artifacts. This paper presents a detailed implementation of YOLOv3 in detecting endoscopic artifacts. Intensive training in a GPU enabled environment (Google COLAB) was carried out. The experimental results of the algorithm achieve mAP% of 55.18 for identifying the artifacts in endoscopic images with a prediction time of 0.0179 seconds on test images.
Keywords: Artifact detection, Deep learning, Endoscopy, YOLOv3.
Scope of the Article: Deep learning