Details
Originalsprache | Englisch |
---|---|
Seiten (von - bis) | 483-492 |
Seitenumfang | 10 |
Fachzeitschrift | International journal of computer assisted radiology and surgery |
Jahrgang | 14 |
Ausgabenummer | 3 |
Frühes Online-Datum | 16 Jan. 2019 |
Publikationsstatus | Veröffentlicht - 14 März 2019 |
Abstract
Purpose: Automated segmentation of anatomical structures in medical image analysis is a prerequisite for autonomous diagnosis as well as various computer- and robot-aided interventions. Recent methods based on deep convolutional neural networks (CNN) have outperformed former heuristic methods. However, those methods were primarily evaluated on rigid, real-world environments. In this study, existing segmentation methods were evaluated for their use on a new dataset of transoral endoscopic exploration. Methods: Four machine learning-based methods SegNet, UNet, ENet and ErfNet were trained with supervision on a novel 7-class dataset of the human larynx. The dataset contains 536 manually segmented images from two patients during laser incisions. The Intersection-over-Union (IoU) evaluation metric was used to measure the accuracy of each method. Data augmentation and network ensembling were employed to increase segmentation accuracy. Stochastic inference was used to show uncertainties of the individual models. Patient-to-patient transfer was investigated using patient-specific fine-tuning. Results: In this study, a weighted average ensemble network of UNet and ErfNet was best suited for the segmentation of laryngeal soft tissue with a mean IoU of 84.7%. The highest efficiency was achieved by ENet with a mean inference time of 9.22 ms per image. It is shown that 10 additional images from a new patient are sufficient for patient-specific fine-tuning. Conclusion: CNN-based methods for semantic segmentation are applicable to endoscopic images of laryngeal soft tissue. The segmentation can be used for active constraints or to monitor morphological changes and autonomously detect pathologies. Further improvements could be achieved by using a larger dataset or training the models in a self-supervised manner on additional unlabeled data.
ASJC Scopus Sachgebiete
- Medizin (insg.)
- Chirurgie
- Ingenieurwesen (insg.)
- Biomedizintechnik
- Medizin (insg.)
- Radiologie, Nuklearmedizin und Bildgebung
- Informatik (insg.)
- Maschinelles Sehen und Mustererkennung
- Medizin (insg.)
- Gesundheitsinformatik
- Informatik (insg.)
- Angewandte Informatik
- Informatik (insg.)
- Computergrafik und computergestütztes Design
Zitieren
- Standard
- Harvard
- Apa
- Vancouver
- BibTex
- RIS
in: International journal of computer assisted radiology and surgery, Jahrgang 14, Nr. 3, 14.03.2019, S. 483-492.
Publikation: Beitrag in Fachzeitschrift › Artikel › Forschung › Peer-Review
}
TY - JOUR
T1 - A dataset of laryngeal endoscopic images with comparative study on convolution neural network-based semantic segmentation
AU - Laves, Max Heinrich
AU - Bicker, Jens
AU - Kahrs, Lüder A.
AU - Ortmaier, Tobias
N1 - Funding information: We thank Giorgio Peretti from the Ospedale Policlinico San Martino, University of Genova, Italy, for providing us with the in vivo laryngeal data used in this study. We would also like to thank James Napier from the Institute of Lasers and Optics, University of Applied Sciences Emden-Leer, Germany, for his thorough proofreading of this manuscript. Funding This research has received funding from the European Union as being part of the ERFE OPhonLas project.
PY - 2019/3/14
Y1 - 2019/3/14
N2 - Purpose: Automated segmentation of anatomical structures in medical image analysis is a prerequisite for autonomous diagnosis as well as various computer- and robot-aided interventions. Recent methods based on deep convolutional neural networks (CNN) have outperformed former heuristic methods. However, those methods were primarily evaluated on rigid, real-world environments. In this study, existing segmentation methods were evaluated for their use on a new dataset of transoral endoscopic exploration. Methods: Four machine learning-based methods SegNet, UNet, ENet and ErfNet were trained with supervision on a novel 7-class dataset of the human larynx. The dataset contains 536 manually segmented images from two patients during laser incisions. The Intersection-over-Union (IoU) evaluation metric was used to measure the accuracy of each method. Data augmentation and network ensembling were employed to increase segmentation accuracy. Stochastic inference was used to show uncertainties of the individual models. Patient-to-patient transfer was investigated using patient-specific fine-tuning. Results: In this study, a weighted average ensemble network of UNet and ErfNet was best suited for the segmentation of laryngeal soft tissue with a mean IoU of 84.7%. The highest efficiency was achieved by ENet with a mean inference time of 9.22 ms per image. It is shown that 10 additional images from a new patient are sufficient for patient-specific fine-tuning. Conclusion: CNN-based methods for semantic segmentation are applicable to endoscopic images of laryngeal soft tissue. The segmentation can be used for active constraints or to monitor morphological changes and autonomously detect pathologies. Further improvements could be achieved by using a larger dataset or training the models in a self-supervised manner on additional unlabeled data.
AB - Purpose: Automated segmentation of anatomical structures in medical image analysis is a prerequisite for autonomous diagnosis as well as various computer- and robot-aided interventions. Recent methods based on deep convolutional neural networks (CNN) have outperformed former heuristic methods. However, those methods were primarily evaluated on rigid, real-world environments. In this study, existing segmentation methods were evaluated for their use on a new dataset of transoral endoscopic exploration. Methods: Four machine learning-based methods SegNet, UNet, ENet and ErfNet were trained with supervision on a novel 7-class dataset of the human larynx. The dataset contains 536 manually segmented images from two patients during laser incisions. The Intersection-over-Union (IoU) evaluation metric was used to measure the accuracy of each method. Data augmentation and network ensembling were employed to increase segmentation accuracy. Stochastic inference was used to show uncertainties of the individual models. Patient-to-patient transfer was investigated using patient-specific fine-tuning. Results: In this study, a weighted average ensemble network of UNet and ErfNet was best suited for the segmentation of laryngeal soft tissue with a mean IoU of 84.7%. The highest efficiency was achieved by ENet with a mean inference time of 9.22 ms per image. It is shown that 10 additional images from a new patient are sufficient for patient-specific fine-tuning. Conclusion: CNN-based methods for semantic segmentation are applicable to endoscopic images of laryngeal soft tissue. The segmentation can be used for active constraints or to monitor morphological changes and autonomously detect pathologies. Further improvements could be achieved by using a larger dataset or training the models in a self-supervised manner on additional unlabeled data.
KW - Computer vision
KW - Larynx
KW - Machine learning
KW - Open-access dataset
KW - Patient-to-patient fine-tuning
KW - Soft tissue
KW - Vocal folds
UR - http://www.scopus.com/inward/record.url?scp=85060178921&partnerID=8YFLogxK
U2 - 10.48550/arXiv.1807.06081
DO - 10.48550/arXiv.1807.06081
M3 - Article
C2 - 30649670
AN - SCOPUS:85060178921
VL - 14
SP - 483
EP - 492
JO - International journal of computer assisted radiology and surgery
JF - International journal of computer assisted radiology and surgery
SN - 1861-6410
IS - 3
ER -