An automated pipeline for constructing 3D models of monogenean hardparts using machine learning technique for landmarks detection / Teo Bee Guan

• Main Author: Teo , Bee Guan
• Format: Thesis
• Published: 2019
• Subjects: Q Science (General)
• Online Access: http://studentsrepo.um.edu.my/11662/1/Teo_Bee_Guan.pdf; http://studentsrepo.um.edu.my/11662/2/Teo_Bee_Guan.pdf; http://studentsrepo.um.edu.my/11662/

Organisms in particular parasitic micro-organisms such as monogeneans defy live investigations due to lack of sophisticated equipment and also due to the difficulties in keeping these delicate organisms alive long enough for investigations to be done. The way monogeneans use their various morphological structures to assist them in their survival have not been studied intensively mainly due to the difficulties in manipulating 2D images and illustrations of attaching structures. Laboratory experiments can only provide a snapshot of a system at any given time but do not fully capture the temporal aspect of the system. Researchers have been looking at alternative methods to assist them in their investigation on how monogeneans function. 3D models may aid researchers in studying morphology and function by rotating the 3D models in 360 degrees for 3D visualization. This task is impossible in 2D illustrations or 2D images. However, the development of 3D models is a tedious procedure as one will have to repeat entire complicated modelling process for every new target 3D shape using a comprehensive 3D modelling software. Hence, this study was designed to develop an alternative 3D modelling approach to build 3D models of monogenean anchors which can be used to understand these morphological structures in three dimensions. The aim of this alternative 3D modelling approach is to avoid repeating the tedious modeling procedure for every single target 3D model from scratch. To achieve this aim, an automated 3D modelling pipeline empowered by an Artificial Neural Network (ANN) was developed. This automated 3D modelling pipeline enables automated deformation of a generic 3D model of monogenean anchor into another target 3D anchor. The entire modelling pipeline can be completed with minimum human intervention by only requiring a 2D illustration of target anchor as an input. The ANN was trained by 5000 synthetic 2D illustrations of monogenean anchors using a data augmentation algorithm. Besides, a NoSQL-based database, MongoDB was developed to store all the synthetic 2D illustrations. The stored illustrations datasets were then retrieved to train the ANN. The proposed 3D modelling pipeline empowered by ANN has managed to automate the generation of the 12 target 3D models (representing 12 species: Cichlidogyrus dracolemma, Dactylogyrus wunderi, Cichlidogyrus raeymaekersi, Cichlidogyrus aspiralis, Dactylogyrus primaries, Pellucidhaptor merus, Dactylogyrus falcatus, Dactylogyrus vastator, Dactylogyrus pterocleidus, Dactylogyrus falciunguis, Chauhanellus auriculatum and Chauhanellus caelatus) of monogenean anchor from the respective 2D illustrations input without repeating the tedious modeling procedure.