A Learning-Guided Hierarchical Approach for Biomedical Image Segmentation

Huaipan Jiang, Anup Sarma, Jihyun Ryoo, Jagadish B. Kotra, Meena Arunachalam, Chitaranjan Das, Mahmut Kandemir

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In recent years, deep convolutional networks have been widely used for a variety of visual recognition tasks, including biomedical applications. In most studies related to biomedical domain (e.g., cell tracking), the first step is to perform symmetric segmentation on target images. Such image datasets have usually the following challenges: (1) they lack human labeled training data, (2) the locations of the objects in images are as equally important as classifying them, and (3) the result accuracy is more critical than that in traditional image segmentation. To address these problems, recent studies employ large deep neural networks to perform segmentation on biomedical images. However, such neural network approaches are very compute intensive due to the high resolution and large quantity of electron microscopy data. Additionally, some of the efforts that make use of neural network models involve redundancy as target biomedical images usually contain smaller regions of interest. Motivated by these observations, in this paper, we propose and experimentally evaluate a more efficient framework, especially suited for image segmentation on embedded systems. This approach involves first 'tiling' the target image, followed by processing the tiles that only contain an object of interest in a hierarchical fashion. Our detailed experimental evaluations using four different datasets indicate that our tiling-based approach can save about 61% of execution time on average, while achieving, at the same time, a slightly higher accuracy compared to the baseline (state of the art) approach.

Original languageEnglish (US)
Title of host publicationProceedings - 31st IEEE International System on Chip Conference, SOCC 2018
EditorsMircea Stan, Karan Bhatia, Helen Li, Ramalingam Sridhar, Massimo Alioto
PublisherIEEE Computer Society
Pages124-129
Number of pages6
ISBN (Electronic)9781538614907
DOIs
StatePublished - Jan 17 2019
Event31st IEEE International System on Chip Conference, SOCC 2018 - Arlington, United States
Duration: Sep 4 2018Sep 7 2018

Publication series

NameInternational System on Chip Conference
Volume2018-September
ISSN (Print)2164-1676
ISSN (Electronic)2164-1706

Conference

Conference31st IEEE International System on Chip Conference, SOCC 2018
CountryUnited States
CityArlington
Period9/4/189/7/18

Fingerprint

Image segmentation
Neural networks
Tile
Embedded systems
Electron microscopy
Redundancy
Processing
Deep neural networks

All Science Journal Classification (ASJC) codes

  • Hardware and Architecture
  • Control and Systems Engineering
  • Electrical and Electronic Engineering

Cite this

Jiang, H., Sarma, A., Ryoo, J., Kotra, J. B., Arunachalam, M., Das, C., & Kandemir, M. (2019). A Learning-Guided Hierarchical Approach for Biomedical Image Segmentation. In M. Stan, K. Bhatia, H. Li, R. Sridhar, & M. Alioto (Eds.), Proceedings - 31st IEEE International System on Chip Conference, SOCC 2018 (pp. 124-129). [8618537] (International System on Chip Conference; Vol. 2018-September). IEEE Computer Society. https://doi.org/10.1109/SOCC.2018.8618537
Jiang, Huaipan ; Sarma, Anup ; Ryoo, Jihyun ; Kotra, Jagadish B. ; Arunachalam, Meena ; Das, Chitaranjan ; Kandemir, Mahmut. / A Learning-Guided Hierarchical Approach for Biomedical Image Segmentation. Proceedings - 31st IEEE International System on Chip Conference, SOCC 2018. editor / Mircea Stan ; Karan Bhatia ; Helen Li ; Ramalingam Sridhar ; Massimo Alioto. IEEE Computer Society, 2019. pp. 124-129 (International System on Chip Conference).
@inproceedings{01cbb89a3bfc4748be618142343da1e9,
title = "A Learning-Guided Hierarchical Approach for Biomedical Image Segmentation",
abstract = "In recent years, deep convolutional networks have been widely used for a variety of visual recognition tasks, including biomedical applications. In most studies related to biomedical domain (e.g., cell tracking), the first step is to perform symmetric segmentation on target images. Such image datasets have usually the following challenges: (1) they lack human labeled training data, (2) the locations of the objects in images are as equally important as classifying them, and (3) the result accuracy is more critical than that in traditional image segmentation. To address these problems, recent studies employ large deep neural networks to perform segmentation on biomedical images. However, such neural network approaches are very compute intensive due to the high resolution and large quantity of electron microscopy data. Additionally, some of the efforts that make use of neural network models involve redundancy as target biomedical images usually contain smaller regions of interest. Motivated by these observations, in this paper, we propose and experimentally evaluate a more efficient framework, especially suited for image segmentation on embedded systems. This approach involves first 'tiling' the target image, followed by processing the tiles that only contain an object of interest in a hierarchical fashion. Our detailed experimental evaluations using four different datasets indicate that our tiling-based approach can save about 61{\%} of execution time on average, while achieving, at the same time, a slightly higher accuracy compared to the baseline (state of the art) approach.",
author = "Huaipan Jiang and Anup Sarma and Jihyun Ryoo and Kotra, {Jagadish B.} and Meena Arunachalam and Chitaranjan Das and Mahmut Kandemir",
year = "2019",
month = "1",
day = "17",
doi = "10.1109/SOCC.2018.8618537",
language = "English (US)",
series = "International System on Chip Conference",
publisher = "IEEE Computer Society",
pages = "124--129",
editor = "Mircea Stan and Karan Bhatia and Helen Li and Ramalingam Sridhar and Massimo Alioto",
booktitle = "Proceedings - 31st IEEE International System on Chip Conference, SOCC 2018",
address = "United States",

}

Jiang, H, Sarma, A, Ryoo, J, Kotra, JB, Arunachalam, M, Das, C & Kandemir, M 2019, A Learning-Guided Hierarchical Approach for Biomedical Image Segmentation. in M Stan, K Bhatia, H Li, R Sridhar & M Alioto (eds), Proceedings - 31st IEEE International System on Chip Conference, SOCC 2018., 8618537, International System on Chip Conference, vol. 2018-September, IEEE Computer Society, pp. 124-129, 31st IEEE International System on Chip Conference, SOCC 2018, Arlington, United States, 9/4/18. https://doi.org/10.1109/SOCC.2018.8618537

A Learning-Guided Hierarchical Approach for Biomedical Image Segmentation. / Jiang, Huaipan; Sarma, Anup; Ryoo, Jihyun; Kotra, Jagadish B.; Arunachalam, Meena; Das, Chitaranjan; Kandemir, Mahmut.

Proceedings - 31st IEEE International System on Chip Conference, SOCC 2018. ed. / Mircea Stan; Karan Bhatia; Helen Li; Ramalingam Sridhar; Massimo Alioto. IEEE Computer Society, 2019. p. 124-129 8618537 (International System on Chip Conference; Vol. 2018-September).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - A Learning-Guided Hierarchical Approach for Biomedical Image Segmentation

AU - Jiang, Huaipan

AU - Sarma, Anup

AU - Ryoo, Jihyun

AU - Kotra, Jagadish B.

AU - Arunachalam, Meena

AU - Das, Chitaranjan

AU - Kandemir, Mahmut

PY - 2019/1/17

Y1 - 2019/1/17

N2 - In recent years, deep convolutional networks have been widely used for a variety of visual recognition tasks, including biomedical applications. In most studies related to biomedical domain (e.g., cell tracking), the first step is to perform symmetric segmentation on target images. Such image datasets have usually the following challenges: (1) they lack human labeled training data, (2) the locations of the objects in images are as equally important as classifying them, and (3) the result accuracy is more critical than that in traditional image segmentation. To address these problems, recent studies employ large deep neural networks to perform segmentation on biomedical images. However, such neural network approaches are very compute intensive due to the high resolution and large quantity of electron microscopy data. Additionally, some of the efforts that make use of neural network models involve redundancy as target biomedical images usually contain smaller regions of interest. Motivated by these observations, in this paper, we propose and experimentally evaluate a more efficient framework, especially suited for image segmentation on embedded systems. This approach involves first 'tiling' the target image, followed by processing the tiles that only contain an object of interest in a hierarchical fashion. Our detailed experimental evaluations using four different datasets indicate that our tiling-based approach can save about 61% of execution time on average, while achieving, at the same time, a slightly higher accuracy compared to the baseline (state of the art) approach.

AB - In recent years, deep convolutional networks have been widely used for a variety of visual recognition tasks, including biomedical applications. In most studies related to biomedical domain (e.g., cell tracking), the first step is to perform symmetric segmentation on target images. Such image datasets have usually the following challenges: (1) they lack human labeled training data, (2) the locations of the objects in images are as equally important as classifying them, and (3) the result accuracy is more critical than that in traditional image segmentation. To address these problems, recent studies employ large deep neural networks to perform segmentation on biomedical images. However, such neural network approaches are very compute intensive due to the high resolution and large quantity of electron microscopy data. Additionally, some of the efforts that make use of neural network models involve redundancy as target biomedical images usually contain smaller regions of interest. Motivated by these observations, in this paper, we propose and experimentally evaluate a more efficient framework, especially suited for image segmentation on embedded systems. This approach involves first 'tiling' the target image, followed by processing the tiles that only contain an object of interest in a hierarchical fashion. Our detailed experimental evaluations using four different datasets indicate that our tiling-based approach can save about 61% of execution time on average, while achieving, at the same time, a slightly higher accuracy compared to the baseline (state of the art) approach.

UR - http://www.scopus.com/inward/record.url?scp=85062244332&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85062244332&partnerID=8YFLogxK

U2 - 10.1109/SOCC.2018.8618537

DO - 10.1109/SOCC.2018.8618537

M3 - Conference contribution

AN - SCOPUS:85062244332

T3 - International System on Chip Conference

SP - 124

EP - 129

BT - Proceedings - 31st IEEE International System on Chip Conference, SOCC 2018

A2 - Stan, Mircea

A2 - Bhatia, Karan

A2 - Li, Helen

A2 - Sridhar, Ramalingam

A2 - Alioto, Massimo

PB - IEEE Computer Society

ER -

Jiang H, Sarma A, Ryoo J, Kotra JB, Arunachalam M, Das C et al. A Learning-Guided Hierarchical Approach for Biomedical Image Segmentation. In Stan M, Bhatia K, Li H, Sridhar R, Alioto M, editors, Proceedings - 31st IEEE International System on Chip Conference, SOCC 2018. IEEE Computer Society. 2019. p. 124-129. 8618537. (International System on Chip Conference). https://doi.org/10.1109/SOCC.2018.8618537