Deep Geodesic Learning for Segmentation and Anatomical Landmarking

Neslisah Torosdagli, Denise K. Liberton, Payal Verma, Murat Sincan, Janice S. Lee, Ulas Bagci

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

In this paper, we propose a novel deep learning framework for anatomy segmentation and automatic landmarking. Specifically, we focus on the challenging problem of mandible segmentation from cone-beam computed tomography (CBCT) scans and identification of 9 anatomical landmarks of the mandible on the geodesic space. The overall approach employs three inter-related steps. In the first step, we propose a deep neural network architecture with carefully designed regularization, and network hyper-parameters to perform image segmentation without the need for data augmentation and complex post-processing refinement. In the second step, we formulate the landmark localization problem directly on the geodesic space for sparsely-spaced anatomical landmarks. In the third step, we utilize a long short-term memory network to identify the closely-spaced landmarks, which is rather difficult to obtain using other standard networks. The proposed fully automated method showed superior efficacy compared to the state-of-the-art mandible segmentation and landmarking approaches in craniofacial anomalies and diseased states. We used a very challenging CBCT data set of 50 patients with a high-degree of craniomaxillofacial variability that is realistic in clinical practice. The qualitative visual inspection was conducted for distinct CBCT scans from 250 patients with high anatomical variability. We have also shown the state-of-the-art performance in an independent data set from the MICCAI Head-Neck Challenge (2015).

Original languageEnglish (US)
Article number8490669
Pages (from-to)919-931
Number of pages13
JournalIEEE Transactions on Medical Imaging
Volume38
Issue number4
DOIs
StatePublished - Apr 2019

Fingerprint

Cone-Beam Computed Tomography
Mandible
Tomography
Cones
Learning
Long-Term Memory
Network architecture
Image segmentation
Short-Term Memory
Anatomy
Neck
Inspection
Head
Processing
Deep learning
Datasets

Keywords

  • Mandible segmentation
  • cone beam computed tomography (CBCT)
  • convolutional neural network
  • craniomaxillofacial deformities
  • deep learning
  • geodesic mapping

ASJC Scopus subject areas

  • Software
  • Radiological and Ultrasound Technology
  • Computer Science Applications
  • Electrical and Electronic Engineering

Cite this

Torosdagli, N., Liberton, D. K., Verma, P., Sincan, M., Lee, J. S., & Bagci, U. (2019). Deep Geodesic Learning for Segmentation and Anatomical Landmarking. IEEE Transactions on Medical Imaging, 38(4), 919-931. [8490669]. https://doi.org/10.1109/TMI.2018.2875814

Deep Geodesic Learning for Segmentation and Anatomical Landmarking. / Torosdagli, Neslisah; Liberton, Denise K.; Verma, Payal; Sincan, Murat; Lee, Janice S.; Bagci, Ulas.

In: IEEE Transactions on Medical Imaging, Vol. 38, No. 4, 8490669, 04.2019, p. 919-931.

Research output: Contribution to journalArticle

Torosdagli, N, Liberton, DK, Verma, P, Sincan, M, Lee, JS & Bagci, U 2019, 'Deep Geodesic Learning for Segmentation and Anatomical Landmarking', IEEE Transactions on Medical Imaging, vol. 38, no. 4, 8490669, pp. 919-931. https://doi.org/10.1109/TMI.2018.2875814
Torosdagli, Neslisah ; Liberton, Denise K. ; Verma, Payal ; Sincan, Murat ; Lee, Janice S. ; Bagci, Ulas. / Deep Geodesic Learning for Segmentation and Anatomical Landmarking. In: IEEE Transactions on Medical Imaging. 2019 ; Vol. 38, No. 4. pp. 919-931.
@article{2286a87e7c36498482d3dc7802e626b7,
title = "Deep Geodesic Learning for Segmentation and Anatomical Landmarking",
abstract = "In this paper, we propose a novel deep learning framework for anatomy segmentation and automatic landmarking. Specifically, we focus on the challenging problem of mandible segmentation from cone-beam computed tomography (CBCT) scans and identification of 9 anatomical landmarks of the mandible on the geodesic space. The overall approach employs three inter-related steps. In the first step, we propose a deep neural network architecture with carefully designed regularization, and network hyper-parameters to perform image segmentation without the need for data augmentation and complex post-processing refinement. In the second step, we formulate the landmark localization problem directly on the geodesic space for sparsely-spaced anatomical landmarks. In the third step, we utilize a long short-term memory network to identify the closely-spaced landmarks, which is rather difficult to obtain using other standard networks. The proposed fully automated method showed superior efficacy compared to the state-of-the-art mandible segmentation and landmarking approaches in craniofacial anomalies and diseased states. We used a very challenging CBCT data set of 50 patients with a high-degree of craniomaxillofacial variability that is realistic in clinical practice. The qualitative visual inspection was conducted for distinct CBCT scans from 250 patients with high anatomical variability. We have also shown the state-of-the-art performance in an independent data set from the MICCAI Head-Neck Challenge (2015).",
keywords = "Mandible segmentation, cone beam computed tomography (CBCT), convolutional neural network, craniomaxillofacial deformities, deep learning, geodesic mapping",
author = "Neslisah Torosdagli and Liberton, {Denise K.} and Payal Verma and Murat Sincan and Lee, {Janice S.} and Ulas Bagci",
year = "2019",
month = "4",
doi = "10.1109/TMI.2018.2875814",
language = "English (US)",
volume = "38",
pages = "919--931",
journal = "IEEE Transactions on Medical Imaging",
issn = "0278-0062",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "4",

}

TY - JOUR

T1 - Deep Geodesic Learning for Segmentation and Anatomical Landmarking

AU - Torosdagli, Neslisah

AU - Liberton, Denise K.

AU - Verma, Payal

AU - Sincan, Murat

AU - Lee, Janice S.

AU - Bagci, Ulas

PY - 2019/4

Y1 - 2019/4

N2 - In this paper, we propose a novel deep learning framework for anatomy segmentation and automatic landmarking. Specifically, we focus on the challenging problem of mandible segmentation from cone-beam computed tomography (CBCT) scans and identification of 9 anatomical landmarks of the mandible on the geodesic space. The overall approach employs three inter-related steps. In the first step, we propose a deep neural network architecture with carefully designed regularization, and network hyper-parameters to perform image segmentation without the need for data augmentation and complex post-processing refinement. In the second step, we formulate the landmark localization problem directly on the geodesic space for sparsely-spaced anatomical landmarks. In the third step, we utilize a long short-term memory network to identify the closely-spaced landmarks, which is rather difficult to obtain using other standard networks. The proposed fully automated method showed superior efficacy compared to the state-of-the-art mandible segmentation and landmarking approaches in craniofacial anomalies and diseased states. We used a very challenging CBCT data set of 50 patients with a high-degree of craniomaxillofacial variability that is realistic in clinical practice. The qualitative visual inspection was conducted for distinct CBCT scans from 250 patients with high anatomical variability. We have also shown the state-of-the-art performance in an independent data set from the MICCAI Head-Neck Challenge (2015).

AB - In this paper, we propose a novel deep learning framework for anatomy segmentation and automatic landmarking. Specifically, we focus on the challenging problem of mandible segmentation from cone-beam computed tomography (CBCT) scans and identification of 9 anatomical landmarks of the mandible on the geodesic space. The overall approach employs three inter-related steps. In the first step, we propose a deep neural network architecture with carefully designed regularization, and network hyper-parameters to perform image segmentation without the need for data augmentation and complex post-processing refinement. In the second step, we formulate the landmark localization problem directly on the geodesic space for sparsely-spaced anatomical landmarks. In the third step, we utilize a long short-term memory network to identify the closely-spaced landmarks, which is rather difficult to obtain using other standard networks. The proposed fully automated method showed superior efficacy compared to the state-of-the-art mandible segmentation and landmarking approaches in craniofacial anomalies and diseased states. We used a very challenging CBCT data set of 50 patients with a high-degree of craniomaxillofacial variability that is realistic in clinical practice. The qualitative visual inspection was conducted for distinct CBCT scans from 250 patients with high anatomical variability. We have also shown the state-of-the-art performance in an independent data set from the MICCAI Head-Neck Challenge (2015).

KW - Mandible segmentation

KW - cone beam computed tomography (CBCT)

KW - convolutional neural network

KW - craniomaxillofacial deformities

KW - deep learning

KW - geodesic mapping

UR - http://www.scopus.com/inward/record.url?scp=85055056033&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85055056033&partnerID=8YFLogxK

U2 - 10.1109/TMI.2018.2875814

DO - 10.1109/TMI.2018.2875814

M3 - Article

C2 - 30334750

AN - SCOPUS:85055056033

VL - 38

SP - 919

EP - 931

JO - IEEE Transactions on Medical Imaging

JF - IEEE Transactions on Medical Imaging

SN - 0278-0062

IS - 4

M1 - 8490669

ER -