Automatic segmentation of human knee anatomy by a convolutional neural network applying a 3D MRI protocol
Abstract Background To study deep learning segmentation of knee anatomy with 13 anatomical classes by using a magnetic resonance (MR) protocol of four three-dimensional (3D) pulse sequences, and evaluate possible clinical usefulness. Methods The sample selection involved 40 healthy right knee volume...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
BMC
2023-01-01
|
| Series: | BMC Musculoskeletal Disorders |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s12891-023-06153-y |
| _version_ | 1797946089387589632 |
|---|---|
| author | Carl Petter Skaar Kulseng Varatharajan Nainamalai Endre Grøvik Jonn-Terje Geitung Asbjørn Årøen Kjell-Inge Gjesdal |
| author_facet | Carl Petter Skaar Kulseng Varatharajan Nainamalai Endre Grøvik Jonn-Terje Geitung Asbjørn Årøen Kjell-Inge Gjesdal |
| author_sort | Carl Petter Skaar Kulseng |
| collection | DOAJ |
| description | Abstract Background To study deep learning segmentation of knee anatomy with 13 anatomical classes by using a magnetic resonance (MR) protocol of four three-dimensional (3D) pulse sequences, and evaluate possible clinical usefulness. Methods The sample selection involved 40 healthy right knee volumes from adult participants. Further, a recently injured single left knee with previous known ACL reconstruction was included as a test subject. The MR protocol consisted of the following 3D pulse sequences: T1 TSE, PD TSE, PD FS TSE, and Angio GE. The DenseVNet neural network was considered for these experiments. Five input combinations of sequences (i) T1, (ii) T1 and FS, (iii) PD and FS, (iv) T1, PD, and FS and (v) T1, PD, FS and Angio were trained using the deep learning algorithm. The Dice similarity coefficient (DSC), Jaccard index and Hausdorff were used to compare the performance of the networks. Results Combining all sequences collectively performed significantly better than other alternatives. The following DSCs (±standard deviation) were obtained for the test dataset: Bone medulla 0.997 (±0.002), PCL 0.973 (±0.015), ACL 0.964 (±0.022), muscle 0.998 (±0.001), cartilage 0.966 (±0.018), bone cortex 0.980 (±0.010), arteries 0.943 (±0.038), collateral ligaments 0.919 (± 0.069), tendons 0.982 (±0.005), meniscus 0.955 (±0.032), adipose tissue 0.998 (±0.001), veins 0.980 (±0.010) and nerves 0.921 (±0.071). The deep learning network correctly identified the anterior cruciate ligament (ACL) tear of the left knee, thus indicating a future aid to orthopaedics. Conclusions The convolutional neural network proves highly capable of correctly labeling all anatomical structures of the knee joint when applied to 3D MR sequences. We have demonstrated that this deep learning model is capable of automatized segmentation that may give 3D models and discover pathology. Both useful for a preoperative evaluation. |
| first_indexed | 2024-04-10T21:05:21Z |
| format | Article |
| id | doaj.art-6d627aa6890f441e83e7d08f5347cf35 |
| institution | Directory Open Access Journal |
| issn | 1471-2474 |
| language | English |
| last_indexed | 2024-04-10T21:05:21Z |
| publishDate | 2023-01-01 |
| publisher | BMC |
| record_format | Article |
| series | BMC Musculoskeletal Disorders |
| spelling | doaj.art-6d627aa6890f441e83e7d08f5347cf352023-01-22T12:02:12ZengBMCBMC Musculoskeletal Disorders1471-24742023-01-0124111210.1186/s12891-023-06153-yAutomatic segmentation of human knee anatomy by a convolutional neural network applying a 3D MRI protocolCarl Petter Skaar Kulseng0Varatharajan Nainamalai1Endre Grøvik2Jonn-Terje Geitung3Asbjørn Årøen4Kjell-Inge Gjesdal5Sunnmøre MR-klinikkNorwegian University of Science and TechnologyNorwegian University of Science and TechnologySunnmøre MR-klinikkDepartment of Orthopedic Surgery, Institute of Clinical Medicine, Akershus University HospitalSunnmøre MR-klinikkAbstract Background To study deep learning segmentation of knee anatomy with 13 anatomical classes by using a magnetic resonance (MR) protocol of four three-dimensional (3D) pulse sequences, and evaluate possible clinical usefulness. Methods The sample selection involved 40 healthy right knee volumes from adult participants. Further, a recently injured single left knee with previous known ACL reconstruction was included as a test subject. The MR protocol consisted of the following 3D pulse sequences: T1 TSE, PD TSE, PD FS TSE, and Angio GE. The DenseVNet neural network was considered for these experiments. Five input combinations of sequences (i) T1, (ii) T1 and FS, (iii) PD and FS, (iv) T1, PD, and FS and (v) T1, PD, FS and Angio were trained using the deep learning algorithm. The Dice similarity coefficient (DSC), Jaccard index and Hausdorff were used to compare the performance of the networks. Results Combining all sequences collectively performed significantly better than other alternatives. The following DSCs (±standard deviation) were obtained for the test dataset: Bone medulla 0.997 (±0.002), PCL 0.973 (±0.015), ACL 0.964 (±0.022), muscle 0.998 (±0.001), cartilage 0.966 (±0.018), bone cortex 0.980 (±0.010), arteries 0.943 (±0.038), collateral ligaments 0.919 (± 0.069), tendons 0.982 (±0.005), meniscus 0.955 (±0.032), adipose tissue 0.998 (±0.001), veins 0.980 (±0.010) and nerves 0.921 (±0.071). The deep learning network correctly identified the anterior cruciate ligament (ACL) tear of the left knee, thus indicating a future aid to orthopaedics. Conclusions The convolutional neural network proves highly capable of correctly labeling all anatomical structures of the knee joint when applied to 3D MR sequences. We have demonstrated that this deep learning model is capable of automatized segmentation that may give 3D models and discover pathology. Both useful for a preoperative evaluation.https://doi.org/10.1186/s12891-023-06153-yMagnetic Resonance ImagingMusculoskeletalDeep learningKnee images segmentationVisualization |
| spellingShingle | Carl Petter Skaar Kulseng Varatharajan Nainamalai Endre Grøvik Jonn-Terje Geitung Asbjørn Årøen Kjell-Inge Gjesdal Automatic segmentation of human knee anatomy by a convolutional neural network applying a 3D MRI protocol BMC Musculoskeletal Disorders Magnetic Resonance Imaging Musculoskeletal Deep learning Knee images segmentation Visualization |
| title | Automatic segmentation of human knee anatomy by a convolutional neural network applying a 3D MRI protocol |
| title_full | Automatic segmentation of human knee anatomy by a convolutional neural network applying a 3D MRI protocol |
| title_fullStr | Automatic segmentation of human knee anatomy by a convolutional neural network applying a 3D MRI protocol |
| title_full_unstemmed | Automatic segmentation of human knee anatomy by a convolutional neural network applying a 3D MRI protocol |
| title_short | Automatic segmentation of human knee anatomy by a convolutional neural network applying a 3D MRI protocol |
| title_sort | automatic segmentation of human knee anatomy by a convolutional neural network applying a 3d mri protocol |
| topic | Magnetic Resonance Imaging Musculoskeletal Deep learning Knee images segmentation Visualization |
| url | https://doi.org/10.1186/s12891-023-06153-y |
| work_keys_str_mv | AT carlpetterskaarkulseng automaticsegmentationofhumankneeanatomybyaconvolutionalneuralnetworkapplyinga3dmriprotocol AT varatharajannainamalai automaticsegmentationofhumankneeanatomybyaconvolutionalneuralnetworkapplyinga3dmriprotocol AT endregrøvik automaticsegmentationofhumankneeanatomybyaconvolutionalneuralnetworkapplyinga3dmriprotocol AT jonnterjegeitung automaticsegmentationofhumankneeanatomybyaconvolutionalneuralnetworkapplyinga3dmriprotocol AT asbjørnarøen automaticsegmentationofhumankneeanatomybyaconvolutionalneuralnetworkapplyinga3dmriprotocol AT kjellingegjesdal automaticsegmentationofhumankneeanatomybyaconvolutionalneuralnetworkapplyinga3dmriprotocol |