TY - GEN
T1 - Cancer detection in mass spectrometry imaging data by dilated convolutional neural networks
AU - van Kersbergen, Jannis
AU - Ghazvinian Zanjani, F.
AU - Zinger, Svitlana
AU - van der Sommen, Fons
AU - Balluff, Benjamin
AU - Vos, D.R.N.
AU - Ellis, S.R.
AU - Heeren, Ron M.A.
AU - Lucas, M.
AU - Marquering, H.A.
AU - Jansen, I.
AU - Savci-Heijink, Cemile Dilara
AU - de Bruin, D.M.
AU - de With, P.H.N.
PY - 2019/3/18
Y1 - 2019/3/18
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%.
KW - Cancer detection
KW - Computational pathology
KW - Convolutional neural networks
KW - Dilated convolution
KW - Mass spectrometry imaging
KW - mass spectrometry imaging
KW - cancer detection
KW - computational pathology
KW - dilated convolution
KW - convolutional neural networks
UR - http://www.scopus.com/inward/record.url?scp=85068686933&partnerID=8YFLogxK
U2 - 10.1117/12.2512360
DO - 10.1117/12.2512360
M3 - Conference contribution
T3 - Proceedings of SPIE
BT - Medical Imaging 2019: Digital Pathology
A2 - Tomaszewski, John E.
A2 - Ward, Aaron D.
PB - SPIE
T2 - SPIE Medical Imaging 2019
Y2 - 16 February 2019 through 21 February 2019
ER -