Implementation of an ideal observer model using convolutional neural network for breast CT images

Gihun Kim, Minah Han, Hyunjung Shim, Jongduk Baek

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

Abstract

In this work, we proposed a non-linear observer model based on convolutional neural network and compare its performance with LG-CHO for four alternative forced choice detection task using simulated breast CT images. In our network, each convolutional layer contained 3×3 filters and a leaky-ReLU as an activation function, but a pooling layer and a zero padding to the output of each convolutional layer were not used unlike general convolutional neural network. Network training was conducted using ADAM optimizer with two design parameters (i.e., network depth and width). The optimal value of the design parameter was found by brute force searching, which spanned up to 30 for depth and 128 for channel, respectively. To generate training and validation dataset, we generated anatomical noise images using a power law spectrum of breast anatomy. 50% volume glandular fraction was assumed, and 1 mm diameter signal was used for detection task. The generated images were recon- structed using filtered back-projection with a fan beam CT geometry, and ramp and Hanning filters were used as an apodization filter to generate different noise structures. To train our network, 125,000 signal present images and 375,000 signal absent images were reconstructed for each apodization filter. To measure detectability, we used percent correction with 4,000 images, generated independently from training and validation dataset. Our results show that the proposed network composed of 30 layers and 64 channels provides higher detectability than LG-CHO. We believe that the improved detectability is achieved by the presence of the non-linear module (i.e., leaky-ReLU) in the network.

Original languageEnglish
Title of host publicationMedical Imaging 2019
Subtitle of host publicationImage Perception, Observer Performance, and Technology Assessment
EditorsRobert M. Nishikawa, Frank W. Samuelson
PublisherSPIE
ISBN (Electronic)9781510625518
DOIs
Publication statusPublished - 2019 Jan 1
EventMedical Imaging 2019: Image Perception, Observer Performance, and Technology Assessment - San Diego, United States
Duration: 2019 Feb 202019 Feb 21

Publication series

NameProgress in Biomedical Optics and Imaging - Proceedings of SPIE
Volume10952
ISSN (Print)1605-7422

Conference

ConferenceMedical Imaging 2019: Image Perception, Observer Performance, and Technology Assessment
CountryUnited States
CitySan Diego
Period19/2/2019/2/21

Fingerprint

breast
Noise
Breast
Neural networks
Architectural Accessibility
Nonlinear Dynamics
Fans
Anatomy
Chemical activation
filters
apodization
education
Geometry
Datasets
anatomy
ramps
fans
modules
projection
activation

All Science Journal Classification (ASJC) codes

  • Electronic, Optical and Magnetic Materials
  • Biomaterials
  • Atomic and Molecular Physics, and Optics
  • Radiology Nuclear Medicine and imaging

Cite this

Kim, G., Han, M., Shim, H., & Baek, J. (2019). Implementation of an ideal observer model using convolutional neural network for breast CT images. In R. M. Nishikawa, & F. W. Samuelson (Eds.), Medical Imaging 2019: Image Perception, Observer Performance, and Technology Assessment [109520L] (Progress in Biomedical Optics and Imaging - Proceedings of SPIE; Vol. 10952). SPIE. https://doi.org/10.1117/12.2512131
Kim, Gihun ; Han, Minah ; Shim, Hyunjung ; Baek, Jongduk. / Implementation of an ideal observer model using convolutional neural network for breast CT images. Medical Imaging 2019: Image Perception, Observer Performance, and Technology Assessment. editor / Robert M. Nishikawa ; Frank W. Samuelson. SPIE, 2019. (Progress in Biomedical Optics and Imaging - Proceedings of SPIE).
@inproceedings{0de1532d785c4836a58081c9acfd79f5,
title = "Implementation of an ideal observer model using convolutional neural network for breast CT images",
abstract = "In this work, we proposed a non-linear observer model based on convolutional neural network and compare its performance with LG-CHO for four alternative forced choice detection task using simulated breast CT images. In our network, each convolutional layer contained 3×3 filters and a leaky-ReLU as an activation function, but a pooling layer and a zero padding to the output of each convolutional layer were not used unlike general convolutional neural network. Network training was conducted using ADAM optimizer with two design parameters (i.e., network depth and width). The optimal value of the design parameter was found by brute force searching, which spanned up to 30 for depth and 128 for channel, respectively. To generate training and validation dataset, we generated anatomical noise images using a power law spectrum of breast anatomy. 50{\%} volume glandular fraction was assumed, and 1 mm diameter signal was used for detection task. The generated images were recon- structed using filtered back-projection with a fan beam CT geometry, and ramp and Hanning filters were used as an apodization filter to generate different noise structures. To train our network, 125,000 signal present images and 375,000 signal absent images were reconstructed for each apodization filter. To measure detectability, we used percent correction with 4,000 images, generated independently from training and validation dataset. Our results show that the proposed network composed of 30 layers and 64 channels provides higher detectability than LG-CHO. We believe that the improved detectability is achieved by the presence of the non-linear module (i.e., leaky-ReLU) in the network.",
author = "Gihun Kim and Minah Han and Hyunjung Shim and Jongduk Baek",
year = "2019",
month = "1",
day = "1",
doi = "10.1117/12.2512131",
language = "English",
series = "Progress in Biomedical Optics and Imaging - Proceedings of SPIE",
publisher = "SPIE",
editor = "Nishikawa, {Robert M.} and Samuelson, {Frank W.}",
booktitle = "Medical Imaging 2019",
address = "United States",

}

Kim, G, Han, M, Shim, H & Baek, J 2019, Implementation of an ideal observer model using convolutional neural network for breast CT images. in RM Nishikawa & FW Samuelson (eds), Medical Imaging 2019: Image Perception, Observer Performance, and Technology Assessment., 109520L, Progress in Biomedical Optics and Imaging - Proceedings of SPIE, vol. 10952, SPIE, Medical Imaging 2019: Image Perception, Observer Performance, and Technology Assessment, San Diego, United States, 19/2/20. https://doi.org/10.1117/12.2512131

Implementation of an ideal observer model using convolutional neural network for breast CT images. / Kim, Gihun; Han, Minah; Shim, Hyunjung; Baek, Jongduk.

Medical Imaging 2019: Image Perception, Observer Performance, and Technology Assessment. ed. / Robert M. Nishikawa; Frank W. Samuelson. SPIE, 2019. 109520L (Progress in Biomedical Optics and Imaging - Proceedings of SPIE; Vol. 10952).

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

TY - GEN

T1 - Implementation of an ideal observer model using convolutional neural network for breast CT images

AU - Kim, Gihun

AU - Han, Minah

AU - Shim, Hyunjung

AU - Baek, Jongduk

PY - 2019/1/1

Y1 - 2019/1/1

N2 - In this work, we proposed a non-linear observer model based on convolutional neural network and compare its performance with LG-CHO for four alternative forced choice detection task using simulated breast CT images. In our network, each convolutional layer contained 3×3 filters and a leaky-ReLU as an activation function, but a pooling layer and a zero padding to the output of each convolutional layer were not used unlike general convolutional neural network. Network training was conducted using ADAM optimizer with two design parameters (i.e., network depth and width). The optimal value of the design parameter was found by brute force searching, which spanned up to 30 for depth and 128 for channel, respectively. To generate training and validation dataset, we generated anatomical noise images using a power law spectrum of breast anatomy. 50% volume glandular fraction was assumed, and 1 mm diameter signal was used for detection task. The generated images were recon- structed using filtered back-projection with a fan beam CT geometry, and ramp and Hanning filters were used as an apodization filter to generate different noise structures. To train our network, 125,000 signal present images and 375,000 signal absent images were reconstructed for each apodization filter. To measure detectability, we used percent correction with 4,000 images, generated independently from training and validation dataset. Our results show that the proposed network composed of 30 layers and 64 channels provides higher detectability than LG-CHO. We believe that the improved detectability is achieved by the presence of the non-linear module (i.e., leaky-ReLU) in the network.

AB - In this work, we proposed a non-linear observer model based on convolutional neural network and compare its performance with LG-CHO for four alternative forced choice detection task using simulated breast CT images. In our network, each convolutional layer contained 3×3 filters and a leaky-ReLU as an activation function, but a pooling layer and a zero padding to the output of each convolutional layer were not used unlike general convolutional neural network. Network training was conducted using ADAM optimizer with two design parameters (i.e., network depth and width). The optimal value of the design parameter was found by brute force searching, which spanned up to 30 for depth and 128 for channel, respectively. To generate training and validation dataset, we generated anatomical noise images using a power law spectrum of breast anatomy. 50% volume glandular fraction was assumed, and 1 mm diameter signal was used for detection task. The generated images were recon- structed using filtered back-projection with a fan beam CT geometry, and ramp and Hanning filters were used as an apodization filter to generate different noise structures. To train our network, 125,000 signal present images and 375,000 signal absent images were reconstructed for each apodization filter. To measure detectability, we used percent correction with 4,000 images, generated independently from training and validation dataset. Our results show that the proposed network composed of 30 layers and 64 channels provides higher detectability than LG-CHO. We believe that the improved detectability is achieved by the presence of the non-linear module (i.e., leaky-ReLU) in the network.

UR - http://www.scopus.com/inward/record.url?scp=85068705059&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85068705059&partnerID=8YFLogxK

U2 - 10.1117/12.2512131

DO - 10.1117/12.2512131

M3 - Conference contribution

AN - SCOPUS:85068705059

T3 - Progress in Biomedical Optics and Imaging - Proceedings of SPIE

BT - Medical Imaging 2019

A2 - Nishikawa, Robert M.

A2 - Samuelson, Frank W.

PB - SPIE

ER -

Kim G, Han M, Shim H, Baek J. Implementation of an ideal observer model using convolutional neural network for breast CT images. In Nishikawa RM, Samuelson FW, editors, Medical Imaging 2019: Image Perception, Observer Performance, and Technology Assessment. SPIE. 2019. 109520L. (Progress in Biomedical Optics and Imaging - Proceedings of SPIE). https://doi.org/10.1117/12.2512131