Automated Knee Bone Segmentation and Visualisation Using Mask RCNN and Marching Cube: Data From The Osteoarthritis Initiative
In this work, an automated knee bone segmentation model is proposed. A mask region-based convolutional neural network (RCNN) algorithm is developed to segment the bone and reconstructed into 3D object by using Marching-Cube algorithm. The proposed method is divided into two stages. First, the Mask R...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Academy of Sciences Malaysia
2022-04-01
|
| Series: | ASM Science Journal |
| Subjects: | |
| Online Access: | https://doi.org/10.32802/asmscj.2022.968 |
| _version_ | 1797900878537031680 |
|---|---|
| author | Rahul Patekar Prashant Shukla Kumar Hong-Seng Gan Muhammad Hanif Ramlee |
| author_facet | Rahul Patekar Prashant Shukla Kumar Hong-Seng Gan Muhammad Hanif Ramlee |
| author_sort | Rahul Patekar |
| collection | DOAJ |
| description | In this work, an automated knee bone segmentation model is proposed. A mask region-based convolutional neural network (RCNN) algorithm is developed to segment the bone and reconstructed into 3D object by using Marching-Cube algorithm. The proposed method is divided into two stages. First, the Mask RCNN is introduced to segment subchondral knee bone from the input MRI sequence. In the second stage, the segmented output from Mask R-CNN is fed as input to the Marching cube algorithm for the 3D reconstruction of knee subchondral bone. The proposed method achieved high dice similarity scores for femur bone 95.35%, tibia bone 95.3%, and patella bone 94.40% using a Mask R-CNN with Resnet-50 as backbone architecture. Improved dice similarity scores for femur bone 97.11%, tibia bone 97.33%, and patella bone 97.05% are obtained by Mask RCNN with Resnet-101 as backbone architecture. It is noted that the Mask RCNN framework has demonstrated efficient and accurate knee subchondral bone detection as well as segmentation for input MRI sequences. |
| first_indexed | 2024-04-10T08:51:56Z |
| format | Article |
| id | doaj.art-09d9c27663e74f8e8497ee5f4df3dc61 |
| institution | Directory Open Access Journal |
| issn | 1823-6782 |
| language | English |
| last_indexed | 2024-04-10T08:51:56Z |
| publishDate | 2022-04-01 |
| publisher | Academy of Sciences Malaysia |
| record_format | Article |
| series | ASM Science Journal |
| spelling | doaj.art-09d9c27663e74f8e8497ee5f4df3dc612023-02-22T04:03:02ZengAcademy of Sciences MalaysiaASM Science Journal1823-67822022-04-01171710.32802/asmscj.2022.968968Automated Knee Bone Segmentation and Visualisation Using Mask RCNN and Marching Cube: Data From The Osteoarthritis InitiativeRahul Patekar0Prashant Shukla Kumar1Hong-Seng Gan2Muhammad Hanif Ramlee3Shri Guru Gobind Singhji Institute of Engineering and TechnologyUniversiti Kuala LumpurUniversiti Kuala LumpurUniversiti Teknologi MalaysiaIn this work, an automated knee bone segmentation model is proposed. A mask region-based convolutional neural network (RCNN) algorithm is developed to segment the bone and reconstructed into 3D object by using Marching-Cube algorithm. The proposed method is divided into two stages. First, the Mask RCNN is introduced to segment subchondral knee bone from the input MRI sequence. In the second stage, the segmented output from Mask R-CNN is fed as input to the Marching cube algorithm for the 3D reconstruction of knee subchondral bone. The proposed method achieved high dice similarity scores for femur bone 95.35%, tibia bone 95.3%, and patella bone 94.40% using a Mask R-CNN with Resnet-50 as backbone architecture. Improved dice similarity scores for femur bone 97.11%, tibia bone 97.33%, and patella bone 97.05% are obtained by Mask RCNN with Resnet-101 as backbone architecture. It is noted that the Mask RCNN framework has demonstrated efficient and accurate knee subchondral bone detection as well as segmentation for input MRI sequences.https://doi.org/10.32802/asmscj.2022.968mask region-based convolutional neural networkosteoarthritismagnetic resonance imagingknee bone segmentation |
| spellingShingle | Rahul Patekar Prashant Shukla Kumar Hong-Seng Gan Muhammad Hanif Ramlee Automated Knee Bone Segmentation and Visualisation Using Mask RCNN and Marching Cube: Data From The Osteoarthritis Initiative ASM Science Journal mask region-based convolutional neural network osteoarthritis magnetic resonance imaging knee bone segmentation |
| title | Automated Knee Bone Segmentation and Visualisation Using Mask RCNN and Marching Cube: Data From The Osteoarthritis Initiative |
| title_full | Automated Knee Bone Segmentation and Visualisation Using Mask RCNN and Marching Cube: Data From The Osteoarthritis Initiative |
| title_fullStr | Automated Knee Bone Segmentation and Visualisation Using Mask RCNN and Marching Cube: Data From The Osteoarthritis Initiative |
| title_full_unstemmed | Automated Knee Bone Segmentation and Visualisation Using Mask RCNN and Marching Cube: Data From The Osteoarthritis Initiative |
| title_short | Automated Knee Bone Segmentation and Visualisation Using Mask RCNN and Marching Cube: Data From The Osteoarthritis Initiative |
| title_sort | automated knee bone segmentation and visualisation using mask rcnn and marching cube data from the osteoarthritis initiative |
| topic | mask region-based convolutional neural network osteoarthritis magnetic resonance imaging knee bone segmentation |
| url | https://doi.org/10.32802/asmscj.2022.968 |
| work_keys_str_mv | AT rahulpatekar automatedkneebonesegmentationandvisualisationusingmaskrcnnandmarchingcubedatafromtheosteoarthritisinitiative AT prashantshuklakumar automatedkneebonesegmentationandvisualisationusingmaskrcnnandmarchingcubedatafromtheosteoarthritisinitiative AT hongsenggan automatedkneebonesegmentationandvisualisationusingmaskrcnnandmarchingcubedatafromtheosteoarthritisinitiative AT muhammadhaniframlee automatedkneebonesegmentationandvisualisationusingmaskrcnnandmarchingcubedatafromtheosteoarthritisinitiative |