Article: Instance segmentation of neuronal cells using U-Net, mask R-CNN, and Detectron2 Journal: International Journal of Biomedical Engineering and Technology (IJBET) 2024 Vol.45 No.2 pp.129 - 149 Abstract: The field of biomedical research has witnessed significant advancements in recent years with the advent of computer vision models for automating cell segmentation. However, the application of such models specifically for neuronal cells is still being determined. Previous studies in computer vision have primarily focused on general cell segmentation, and only a few have investigated the application of specific models for neuronal cells. This gap in research inspired the current study to investigate the application and evaluation of U-Net, mask region-based convolutional neural network (mask R-CNN), and Detectron2 models for automating the cell segmentation of microscopic images of neuronal cells. The study utilised three evaluation metrics, including intersection over union, size, and robustness, to determine the accuracy and efficiency of the models. Results showed that Detectron2 outperformed the other two models, demonstrating higher accuracy and robustness in neuronal cell segmentation, enabling new insights and applications in biomedicine through efficient evaluation of neurological disorders treatments. Inderscience Publishers - linking academia, business and industry through research

Title: Instance segmentation of neuronal cells using U-Net, mask R-CNN, and Detectron2

Authors: Ramasamy Murugesan; Gokul Adethya T.; Ventesh A.; B. Azhaganathan; P.D.D. Domnic

Addresses: National Institute of Technology, Tiruchirappalli, Tamil Nadu, 620015, India ' National Institute of Technology, Tiruchirappalli, Tamil Nadu, 620015, India ' National Institute of Technology, Tiruchirappalli, Tamil Nadu, 620015, India ' National Institute of Technology, Tiruchirappalli, Tamil Nadu, 620015, India ' Universiti Teknologi Petronas, Seri Iskandar, Malaysia

Abstract: The field of biomedical research has witnessed significant advancements in recent years with the advent of computer vision models for automating cell segmentation. However, the application of such models specifically for neuronal cells is still being determined. Previous studies in computer vision have primarily focused on general cell segmentation, and only a few have investigated the application of specific models for neuronal cells. This gap in research inspired the current study to investigate the application and evaluation of U-Net, mask region-based convolutional neural network (mask R-CNN), and Detectron2 models for automating the cell segmentation of microscopic images of neuronal cells. The study utilised three evaluation metrics, including intersection over union, size, and robustness, to determine the accuracy and efficiency of the models. Results showed that Detectron2 outperformed the other two models, demonstrating higher accuracy and robustness in neuronal cell segmentation, enabling new insights and applications in biomedicine through efficient evaluation of neurological disorders treatments.

Keywords: deep learning; computer vision; cell instance segmentation; U-Net; mask R-CNN; Detectron2.

DOI: 10.1504/IJBET.2024.138735

International Journal of Biomedical Engineering and Technology, 2024 Vol.45 No.2, pp.129 - 149

Received: 19 May 2023
Accepted: 16 Sep 2023
Published online: 29 May 2024 *

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

Title: Instance segmentation of neuronal cells using U-Net, mask R-CNN, and Detectron2

Keep up-to-date