Perceiving Systems Conference Paper 2020

GENTEL: GENerating Training data Efficiently for Learning to segment medical images

Gentel

Accurately segmenting MRI images is crucial for many clinical applications. However, manually segmenting images with accurate pixel precision is a tedious and time consuming task. In this paper we present a simple, yet effective method to improve the efficiency of the image segmentation process. We propose to transform the image annotation task into a binary choice task. We start by using classical image processing algorithms with different parameter values to generate multiple, different segmentation masks for each input MRI image. Then, instead of segmenting the pixels of the images, the user only needs to decide whether a segmentation is acceptable or not. This method allows us to efficiently obtain high quality segmentations with minor human intervention. With the selected segmentations, we train a state-of-the-art neural network model. For the evaluation, we use a second MRI dataset (1.5T Dataset), acquired with a different protocol and containing annotations. We show that the trained network i) is able to automatically segment cases where none of the classical methods obtain a high quality result ; ii) generalizes to the second MRI dataset, which was acquired with a different protocol and was never seen at training time ; and iii) enables detection of miss-annotations in this second dataset. Quantitatively, the trained network obtains very good results: DICE score - mean 0.98, median 0.99- and Hausdorff distance (in pixels) - mean 4.7, median 2.0-.

Author(s): Rajat Prince Thakur and Sergi Pujades and Lavika Goel and Rolf Pohmann and Jürgen Machann and Michael J. Black
Book Title: RFIAP 2020 - Congrés Reconnaissance des Formes, Image, Apprentissage et Perception
Year: 2020
Month: June
Project(s):
Bibtex Type: Conference Paper (conference)
Event Name: Congrès Reconnaissance des Formes, Image, Apprentissage et Perception (RFAIP 2020)
Event Place: Vannes, France
State: Published
URL: https://inria.hal.science/hal-02988367/
Electronic Archiving: grant_archive
Links:
Attachments:

BibTex

@conference{gentel:rfiap:2020,
  title = {{GENTEL: GENerating Training data Efficiently for Learning to segment medical images}},
  booktitle = {RFIAP 2020 - Congrés Reconnaissance des Formes, Image, Apprentissage et Perception },
  abstract = {Accurately segmenting MRI images is crucial for many clinical applications. However, manually segmenting images with accurate pixel precision is a tedious and time consuming task. In this paper we present a simple, yet effective method to improve the efficiency of the image segmentation process. We propose to transform the image annotation task into a binary choice task. We start by using classical image processing algorithms with different parameter values to generate multiple, different segmentation masks for each input MRI image. Then, instead of segmenting the pixels of the images, the user only needs to decide whether a segmentation is acceptable or not. This method allows us to efficiently obtain high quality segmentations with minor human intervention. With the selected segmentations, we train a state-of-the-art neural network model. For the evaluation, we use a second MRI dataset (1.5T Dataset), acquired with a different protocol and containing annotations. We show that the trained network i) is able to automatically segment cases where none of the classical methods obtain a high quality result ; ii) generalizes to the second MRI dataset, which was acquired with a different protocol and was never seen at training time ; and iii) enables detection of miss-annotations in this second dataset. Quantitatively, the trained network obtains very good results: DICE score - mean 0.98, median 0.99- and Hausdorff distance (in pixels) - mean 4.7, median 2.0-.},
  month = jun,
  year = {2020},
  slug = {gentel-rfiap-2020},
  author = {Thakur, Rajat Prince and Pujades, Sergi and Goel, Lavika and Pohmann, Rolf and Machann, J\"{u}rgen and Black, Michael J.},
  url = {https://inria.hal.science/hal-02988367/},
  month_numeric = {6}
}