Cancer detection in mass spectrometry imaging data by dilated convolutional neural networks

J. van Kersbergen, F. Ghazvinian Zanjani, S. Zinger, F. van der Sommen, B. Balluff, D. R. N. Vos, S. R. Ellis, R. M. A. Heeran, M. Lucas, H. A. Marquering, I. Jansen, C. D. Savci-Heijink, D. M. de Bruin, P. H. N. de With

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

Abstract

Imaging mass spectrometry (IMS) is a novel molecular imaging technique to investigate how molecules are distributed between tumors and within tumor region in order to shed light into tumor biology or find potential biomarkers. Convolutional neural networks (CNNs) have proven to be very potent classifiers often outperforming other machine learning algorithms, especially in computational pathology. To overcome the challenge of complexity and high-dimensionality of the IMS data, the proposed CNNs are either very deep or use large kernels, which results in large amount of parameters and therefore a high computational complexity. An alternative is down-sampling the data, which inherently leads to a loss of information. In this paper, we propose using dilated CNNs as a possible solution to this challenge, since it allows for an increase of the receptive field size, neither by increasing the network parameters nor by decreasing the input signal resolution. Since the mass signature of cancer biomarkers are distributed over the whole mass spectrum, both locally-and globally-distributed patterns need to be captured to correctly classify the spectrum. By experiment, we show that employing dilated convolutions in the architecture of a CNN leads to a higher performance in tumor classification. Our proposed model outperforms the state-of-the-art for tumor classification in both clinical lung and bladder datasets by 1-3%.
Original languageEnglish
Title of host publicationMedical Imaging 2019: Digital Pathology
EditorsJohn E. Tomaszewski, Aaron D. Ward
PublisherSPIE
Volume10956
ISBN (Electronic)9781510625594
DOIs
Publication statusPublished - 2019
EventMedical Imaging 2019: Digital Pathology - San Diego, United States
Duration: 20 Feb 201921 Feb 2019

Publication series

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

Conference

ConferenceMedical Imaging 2019: Digital Pathology
CountryUnited States
CitySan Diego
Period20/02/201921/02/2019

Cite this

van Kersbergen, J., Ghazvinian Zanjani, F., Zinger, S., van der Sommen, F., Balluff, B., Vos, D. R. N., ... de With, P. H. N. (2019). Cancer detection in mass spectrometry imaging data by dilated convolutional neural networks. In J. E. Tomaszewski, & A. D. Ward (Eds.), Medical Imaging 2019: Digital Pathology (Vol. 10956). [109560I] (Progress in Biomedical Optics and Imaging - Proceedings of SPIE). SPIE. https://doi.org/10.1117/12.2512360
van Kersbergen, J. ; Ghazvinian Zanjani, F. ; Zinger, S. ; van der Sommen, F. ; Balluff, B. ; Vos, D. R. N. ; Ellis, S. R. ; Heeran, R. M. A. ; Lucas, M. ; Marquering, H. A. ; Jansen, I. ; Savci-Heijink, C. D. ; de Bruin, D. M. ; de With, P. H. N. / Cancer detection in mass spectrometry imaging data by dilated convolutional neural networks. Medical Imaging 2019: Digital Pathology. editor / John E. Tomaszewski ; Aaron D. Ward. Vol. 10956 SPIE, 2019. (Progress in Biomedical Optics and Imaging - Proceedings of SPIE).
@inproceedings{6cf5bb4182b24b689ec9528dc8f8a2b4,
title = "Cancer detection in mass spectrometry imaging data by dilated convolutional neural networks",
abstract = "Imaging mass spectrometry (IMS) is a novel molecular imaging technique to investigate how molecules are distributed between tumors and within tumor region in order to shed light into tumor biology or find potential biomarkers. Convolutional neural networks (CNNs) have proven to be very potent classifiers often outperforming other machine learning algorithms, especially in computational pathology. To overcome the challenge of complexity and high-dimensionality of the IMS data, the proposed CNNs are either very deep or use large kernels, which results in large amount of parameters and therefore a high computational complexity. An alternative is down-sampling the data, which inherently leads to a loss of information. In this paper, we propose using dilated CNNs as a possible solution to this challenge, since it allows for an increase of the receptive field size, neither by increasing the network parameters nor by decreasing the input signal resolution. Since the mass signature of cancer biomarkers are distributed over the whole mass spectrum, both locally-and globally-distributed patterns need to be captured to correctly classify the spectrum. By experiment, we show that employing dilated convolutions in the architecture of a CNN leads to a higher performance in tumor classification. Our proposed model outperforms the state-of-the-art for tumor classification in both clinical lung and bladder datasets by 1-3{\%}.",
author = "{van Kersbergen}, J. and {Ghazvinian Zanjani}, F. and S. Zinger and {van der Sommen}, F. and B. Balluff and Vos, {D. R. N.} and Ellis, {S. R.} and Heeran, {R. M. A.} and M. Lucas and Marquering, {H. A.} and I. Jansen and Savci-Heijink, {C. D.} and {de Bruin}, {D. M.} and {de With}, {P. H. N.}",
year = "2019",
doi = "10.1117/12.2512360",
language = "English",
volume = "10956",
series = "Progress in Biomedical Optics and Imaging - Proceedings of SPIE",
publisher = "SPIE",
editor = "Tomaszewski, {John E.} and Ward, {Aaron D.}",
booktitle = "Medical Imaging 2019: Digital Pathology",
address = "United States",

}

van Kersbergen, J, Ghazvinian Zanjani, F, Zinger, S, van der Sommen, F, Balluff, B, Vos, DRN, Ellis, SR, Heeran, RMA, Lucas, M, Marquering, HA, Jansen, I, Savci-Heijink, CD, de Bruin, DM & de With, PHN 2019, Cancer detection in mass spectrometry imaging data by dilated convolutional neural networks. in JE Tomaszewski & AD Ward (eds), Medical Imaging 2019: Digital Pathology. vol. 10956, 109560I, Progress in Biomedical Optics and Imaging - Proceedings of SPIE, SPIE, Medical Imaging 2019: Digital Pathology, San Diego, United States, 20/02/2019. https://doi.org/10.1117/12.2512360

Cancer detection in mass spectrometry imaging data by dilated convolutional neural networks. / van Kersbergen, J.; Ghazvinian Zanjani, F.; Zinger, S.; van der Sommen, F.; Balluff, B.; Vos, D. R. N.; Ellis, S. R.; Heeran, R. M. A.; Lucas, M.; Marquering, H. A.; Jansen, I.; Savci-Heijink, C. D.; de Bruin, D. M.; de With, P. H. N.

Medical Imaging 2019: Digital Pathology. ed. / John E. Tomaszewski; Aaron D. Ward. Vol. 10956 SPIE, 2019. 109560I (Progress in Biomedical Optics and Imaging - Proceedings of SPIE).

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - Cancer detection in mass spectrometry imaging data by dilated convolutional neural networks

AU - van Kersbergen, J.

AU - Ghazvinian Zanjani, F.

AU - Zinger, S.

AU - van der Sommen, F.

AU - Balluff, B.

AU - Vos, D. R. N.

AU - Ellis, S. R.

AU - Heeran, R. M. A.

AU - Lucas, M.

AU - Marquering, H. A.

AU - Jansen, I.

AU - Savci-Heijink, C. D.

AU - de Bruin, D. M.

AU - de With, P. H. N.

PY - 2019

Y1 - 2019

N2 - Imaging mass spectrometry (IMS) is a novel molecular imaging technique to investigate how molecules are distributed between tumors and within tumor region in order to shed light into tumor biology or find potential biomarkers. Convolutional neural networks (CNNs) have proven to be very potent classifiers often outperforming other machine learning algorithms, especially in computational pathology. To overcome the challenge of complexity and high-dimensionality of the IMS data, the proposed CNNs are either very deep or use large kernels, which results in large amount of parameters and therefore a high computational complexity. An alternative is down-sampling the data, which inherently leads to a loss of information. In this paper, we propose using dilated CNNs as a possible solution to this challenge, since it allows for an increase of the receptive field size, neither by increasing the network parameters nor by decreasing the input signal resolution. Since the mass signature of cancer biomarkers are distributed over the whole mass spectrum, both locally-and globally-distributed patterns need to be captured to correctly classify the spectrum. By experiment, we show that employing dilated convolutions in the architecture of a CNN leads to a higher performance in tumor classification. Our proposed model outperforms the state-of-the-art for tumor classification in both clinical lung and bladder datasets by 1-3%.

AB - Imaging mass spectrometry (IMS) is a novel molecular imaging technique to investigate how molecules are distributed between tumors and within tumor region in order to shed light into tumor biology or find potential biomarkers. Convolutional neural networks (CNNs) have proven to be very potent classifiers often outperforming other machine learning algorithms, especially in computational pathology. To overcome the challenge of complexity and high-dimensionality of the IMS data, the proposed CNNs are either very deep or use large kernels, which results in large amount of parameters and therefore a high computational complexity. An alternative is down-sampling the data, which inherently leads to a loss of information. In this paper, we propose using dilated CNNs as a possible solution to this challenge, since it allows for an increase of the receptive field size, neither by increasing the network parameters nor by decreasing the input signal resolution. Since the mass signature of cancer biomarkers are distributed over the whole mass spectrum, both locally-and globally-distributed patterns need to be captured to correctly classify the spectrum. By experiment, we show that employing dilated convolutions in the architecture of a CNN leads to a higher performance in tumor classification. Our proposed model outperforms the state-of-the-art for tumor classification in both clinical lung and bladder datasets by 1-3%.

UR - https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85068686933&origin=inward

U2 - 10.1117/12.2512360

DO - 10.1117/12.2512360

M3 - Conference contribution

VL - 10956

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

BT - Medical Imaging 2019: Digital Pathology

A2 - Tomaszewski, John E.

A2 - Ward, Aaron D.

PB - SPIE

ER -

van Kersbergen J, Ghazvinian Zanjani F, Zinger S, van der Sommen F, Balluff B, Vos DRN et al. Cancer detection in mass spectrometry imaging data by dilated convolutional neural networks. In Tomaszewski JE, Ward AD, editors, Medical Imaging 2019: Digital Pathology. Vol. 10956. SPIE. 2019. 109560I. (Progress in Biomedical Optics and Imaging - Proceedings of SPIE). https://doi.org/10.1117/12.2512360