Malignant microcalcification clusters detection using unsupervised deep autoencoders

Rui Hou, Yinhao Ren, Lars J. Grimm, MacIej A. Mazurowski, Jeffrey R. Marks, Lorraine King, Carlo Maley, E. Shelley Hwang, Joseph Y. Lo

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

Abstract

Detection and localization of microcalcification (MC) clusters are very important in mammography diagnosis. Supervised MC detectors require learning from extracted individual MCs and MC clusters. However, they are limited by number of datasets given that MC images are hard to obtain. In this work, we propose a method to detect malignant microcalcification (MC) clusters using unsupervised, one-class, deep convolutional autoencoder. Specifically, we designed a deep autoencoder model where only patches extracted from normal cases' mammograms are used during training. We then applied our trained model on patches extracted from testing images. Our training dataset contains 408 normal subjects, including 1961 full-field digital mammography images. Our testing datasets contains 276 subjects. Specifically, 106 of them were patients diagnosed with Ductal Carcinoma In-Situ (DCIS); 70 of them were diagnosed with Invasive Ductal Carcinoma (IDC); the rest 100 are normal cases containing 484 negative screening mammograms. Patches extracted from DCIS and IDC cases (positive patches) contain MC clusters, whereas patches extracted from normal cases (negative patches) don't. As the model is trained only on negative images that do not contain MCs, it cannot reconstruct MCs well, and thus, the reconstruction error will be larger on positive patches than negative patches. Our detection algorithm's decision is made based on Max-Squared Error between autoencoder's input and output patches. To confirm the results were not simply due to blurring, we then compared our designed detector with unsharp mask with Gaussian blur results. The results using the unsupervised autoencoder on testing patches with size 64×64 achieves an AUC result of 0.93. The best performance on testing patches using Gaussian blur with kernel size equal to 11has an overall AUC of 0.82.

Original languageEnglish (US)
Title of host publicationMedical Imaging 2019
Subtitle of host publicationComputer-Aided Diagnosis
EditorsKensaku Mori, Horst K. Hahn
PublisherSPIE
ISBN (Electronic)9781510625471
DOIs
StatePublished - Jan 1 2019
EventMedical Imaging 2019: Computer-Aided Diagnosis - San Diego, United States
Duration: Feb 17 2019Feb 20 2019

Publication series

NameProgress in Biomedical Optics and Imaging - Proceedings of SPIE
Volume10950
ISSN (Print)1605-7422

Conference

ConferenceMedical Imaging 2019: Computer-Aided Diagnosis
CountryUnited States
CitySan Diego
Period2/17/192/20/19

Fingerprint

Calcinosis
cancer
Mammography
Testing
Ductal Carcinoma
Carcinoma, Intraductal, Noninfiltrating
education
Detectors
Area Under Curve
blurring
detectors
learning
Masks
Screening
masks
screening
output
Learning
Datasets

Keywords

  • Autoencoder
  • DCIS
  • IDC
  • Malignant
  • Mammogram
  • Microcalcifications
  • Unsupervised learning

ASJC Scopus subject areas

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

Cite this

Hou, R., Ren, Y., Grimm, L. J., Mazurowski, M. A., Marks, J. R., King, L., ... Lo, J. Y. (2019). Malignant microcalcification clusters detection using unsupervised deep autoencoders. In K. Mori, & H. K. Hahn (Eds.), Medical Imaging 2019: Computer-Aided Diagnosis [109502Q] (Progress in Biomedical Optics and Imaging - Proceedings of SPIE; Vol. 10950). SPIE. https://doi.org/10.1117/12.2512829

Malignant microcalcification clusters detection using unsupervised deep autoencoders. / Hou, Rui; Ren, Yinhao; Grimm, Lars J.; Mazurowski, MacIej A.; Marks, Jeffrey R.; King, Lorraine; Maley, Carlo; Shelley Hwang, E.; Lo, Joseph Y.

Medical Imaging 2019: Computer-Aided Diagnosis. ed. / Kensaku Mori; Horst K. Hahn. SPIE, 2019. 109502Q (Progress in Biomedical Optics and Imaging - Proceedings of SPIE; Vol. 10950).

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

Hou, R, Ren, Y, Grimm, LJ, Mazurowski, MA, Marks, JR, King, L, Maley, C, Shelley Hwang, E & Lo, JY 2019, Malignant microcalcification clusters detection using unsupervised deep autoencoders. in K Mori & HK Hahn (eds), Medical Imaging 2019: Computer-Aided Diagnosis., 109502Q, Progress in Biomedical Optics and Imaging - Proceedings of SPIE, vol. 10950, SPIE, Medical Imaging 2019: Computer-Aided Diagnosis, San Diego, United States, 2/17/19. https://doi.org/10.1117/12.2512829
Hou R, Ren Y, Grimm LJ, Mazurowski MA, Marks JR, King L et al. Malignant microcalcification clusters detection using unsupervised deep autoencoders. In Mori K, Hahn HK, editors, Medical Imaging 2019: Computer-Aided Diagnosis. SPIE. 2019. 109502Q. (Progress in Biomedical Optics and Imaging - Proceedings of SPIE). https://doi.org/10.1117/12.2512829
Hou, Rui ; Ren, Yinhao ; Grimm, Lars J. ; Mazurowski, MacIej A. ; Marks, Jeffrey R. ; King, Lorraine ; Maley, Carlo ; Shelley Hwang, E. ; Lo, Joseph Y. / Malignant microcalcification clusters detection using unsupervised deep autoencoders. Medical Imaging 2019: Computer-Aided Diagnosis. editor / Kensaku Mori ; Horst K. Hahn. SPIE, 2019. (Progress in Biomedical Optics and Imaging - Proceedings of SPIE).
@inproceedings{c9865927933c4d47a5554598b366c15c,
title = "Malignant microcalcification clusters detection using unsupervised deep autoencoders",
abstract = "Detection and localization of microcalcification (MC) clusters are very important in mammography diagnosis. Supervised MC detectors require learning from extracted individual MCs and MC clusters. However, they are limited by number of datasets given that MC images are hard to obtain. In this work, we propose a method to detect malignant microcalcification (MC) clusters using unsupervised, one-class, deep convolutional autoencoder. Specifically, we designed a deep autoencoder model where only patches extracted from normal cases' mammograms are used during training. We then applied our trained model on patches extracted from testing images. Our training dataset contains 408 normal subjects, including 1961 full-field digital mammography images. Our testing datasets contains 276 subjects. Specifically, 106 of them were patients diagnosed with Ductal Carcinoma In-Situ (DCIS); 70 of them were diagnosed with Invasive Ductal Carcinoma (IDC); the rest 100 are normal cases containing 484 negative screening mammograms. Patches extracted from DCIS and IDC cases (positive patches) contain MC clusters, whereas patches extracted from normal cases (negative patches) don't. As the model is trained only on negative images that do not contain MCs, it cannot reconstruct MCs well, and thus, the reconstruction error will be larger on positive patches than negative patches. Our detection algorithm's decision is made based on Max-Squared Error between autoencoder's input and output patches. To confirm the results were not simply due to blurring, we then compared our designed detector with unsharp mask with Gaussian blur results. The results using the unsupervised autoencoder on testing patches with size 64×64 achieves an AUC result of 0.93. The best performance on testing patches using Gaussian blur with kernel size equal to 11has an overall AUC of 0.82.",
keywords = "Autoencoder, DCIS, IDC, Malignant, Mammogram, Microcalcifications, Unsupervised learning",
author = "Rui Hou and Yinhao Ren and Grimm, {Lars J.} and Mazurowski, {MacIej A.} and Marks, {Jeffrey R.} and Lorraine King and Carlo Maley and {Shelley Hwang}, E. and Lo, {Joseph Y.}",
year = "2019",
month = "1",
day = "1",
doi = "10.1117/12.2512829",
language = "English (US)",
series = "Progress in Biomedical Optics and Imaging - Proceedings of SPIE",
publisher = "SPIE",
editor = "Kensaku Mori and Hahn, {Horst K.}",
booktitle = "Medical Imaging 2019",

}

TY - GEN

T1 - Malignant microcalcification clusters detection using unsupervised deep autoencoders

AU - Hou, Rui

AU - Ren, Yinhao

AU - Grimm, Lars J.

AU - Mazurowski, MacIej A.

AU - Marks, Jeffrey R.

AU - King, Lorraine

AU - Maley, Carlo

AU - Shelley Hwang, E.

AU - Lo, Joseph Y.

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Detection and localization of microcalcification (MC) clusters are very important in mammography diagnosis. Supervised MC detectors require learning from extracted individual MCs and MC clusters. However, they are limited by number of datasets given that MC images are hard to obtain. In this work, we propose a method to detect malignant microcalcification (MC) clusters using unsupervised, one-class, deep convolutional autoencoder. Specifically, we designed a deep autoencoder model where only patches extracted from normal cases' mammograms are used during training. We then applied our trained model on patches extracted from testing images. Our training dataset contains 408 normal subjects, including 1961 full-field digital mammography images. Our testing datasets contains 276 subjects. Specifically, 106 of them were patients diagnosed with Ductal Carcinoma In-Situ (DCIS); 70 of them were diagnosed with Invasive Ductal Carcinoma (IDC); the rest 100 are normal cases containing 484 negative screening mammograms. Patches extracted from DCIS and IDC cases (positive patches) contain MC clusters, whereas patches extracted from normal cases (negative patches) don't. As the model is trained only on negative images that do not contain MCs, it cannot reconstruct MCs well, and thus, the reconstruction error will be larger on positive patches than negative patches. Our detection algorithm's decision is made based on Max-Squared Error between autoencoder's input and output patches. To confirm the results were not simply due to blurring, we then compared our designed detector with unsharp mask with Gaussian blur results. The results using the unsupervised autoencoder on testing patches with size 64×64 achieves an AUC result of 0.93. The best performance on testing patches using Gaussian blur with kernel size equal to 11has an overall AUC of 0.82.

AB - Detection and localization of microcalcification (MC) clusters are very important in mammography diagnosis. Supervised MC detectors require learning from extracted individual MCs and MC clusters. However, they are limited by number of datasets given that MC images are hard to obtain. In this work, we propose a method to detect malignant microcalcification (MC) clusters using unsupervised, one-class, deep convolutional autoencoder. Specifically, we designed a deep autoencoder model where only patches extracted from normal cases' mammograms are used during training. We then applied our trained model on patches extracted from testing images. Our training dataset contains 408 normal subjects, including 1961 full-field digital mammography images. Our testing datasets contains 276 subjects. Specifically, 106 of them were patients diagnosed with Ductal Carcinoma In-Situ (DCIS); 70 of them were diagnosed with Invasive Ductal Carcinoma (IDC); the rest 100 are normal cases containing 484 negative screening mammograms. Patches extracted from DCIS and IDC cases (positive patches) contain MC clusters, whereas patches extracted from normal cases (negative patches) don't. As the model is trained only on negative images that do not contain MCs, it cannot reconstruct MCs well, and thus, the reconstruction error will be larger on positive patches than negative patches. Our detection algorithm's decision is made based on Max-Squared Error between autoencoder's input and output patches. To confirm the results were not simply due to blurring, we then compared our designed detector with unsharp mask with Gaussian blur results. The results using the unsupervised autoencoder on testing patches with size 64×64 achieves an AUC result of 0.93. The best performance on testing patches using Gaussian blur with kernel size equal to 11has an overall AUC of 0.82.

KW - Autoencoder

KW - DCIS

KW - IDC

KW - Malignant

KW - Mammogram

KW - Microcalcifications

KW - Unsupervised learning

UR - http://www.scopus.com/inward/record.url?scp=85068175819&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85068175819&partnerID=8YFLogxK

U2 - 10.1117/12.2512829

DO - 10.1117/12.2512829

M3 - Conference contribution

AN - SCOPUS:85068175819

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

BT - Medical Imaging 2019

A2 - Mori, Kensaku

A2 - Hahn, Horst K.

PB - SPIE

ER -