|
Towards Autonomous Micropipette Positioning in Eye Surgery by Employing Deep Learning Algorithm in Micro-Cannulation |
|---|---|
| รหัสดีโอไอ | |
| Creator | Mukesh Madanan, Nurul Akhmal Mohd Zulkefli |
| Title | Towards Autonomous Micropipette Positioning in Eye Surgery by Employing Deep Learning Algorithm in Micro-Cannulation |
| Contributor | - |
| Publisher | TuEngr Group |
| Publication Year | 2566 |
| Journal Title | International Transaction Journal of Engineering, Management, & Applied Sciences & Technologies |
| Journal Vol. | 14 |
| Journal No. | 1 |
| Page no. | 14A1A: 1-20 |
| Keyword | Artificial Intelligence, Machine Learning, Deep Learning, Robotic Surgery, Eye Surgery, Micro-cannulation, Enhanced Guassian Filtering, Bee Colony Optimization, CNN, Image Processing |
| URL Website | http://TuEngr.com/Vol14-1.html |
| Website title | ITJEMAST V14(1) 2023 @ TuEngr.com |
| ISSN | 2228-9860 |
| Abstract | Eye surgery, more precisely the retinal micro-surgery involves both sensory as well as motor skills. This is confined within human boundaries along with physiological limits for maintaining consistent steadiness, the ability to feel small forces and accuracy. Despite these assumptions to leverage robots in all types of surgery, multitudes of challenges have to be confronted to reach complete development. The deployment of robotic assistance in ophthalmologic surgery also faces the same challenge. This work focuses on the autonomous positioning of a micropipette that is to be mounted on a surgical robot for performing eye surgery. Initially, multiple microscopic images of the given micropipette along with its shadow are collected. These images are treated or filtered by using the Enhanced Gaussian Filtering (EGF) method. The so-obtained filtered image is partitioned or segmented by Bee Colony Optimization (BCO) into three segments: micropipette, eye ground and shadow of the micropipette. A new Modified Convolutional Neural Network (MCNN) is leveraged by the robot to perform eye surgery that learns the microscopic images with their ground truth. This MCNN uses automatic feature extraction and estimates micropipette regions with their shadow by examining a microscopic image and its tip. This is tapped for developing autonomous position control in robots. The selected micropipette is found to be positioned at a 99.56% success rate with a mean distance of 1.37 mm from the eye ground that is simulated. |
