Convolutional neural networks for segmentation and object detection of human semen

Malte Stær Nissen; Oswin Krause; Kristian Almstrup; Søren Kjærulff; Torben T. Nielsen; Mads Nielsen

doi:10.1007/978-3-319-59126-1_33

Convolutional neural networks for segmentation and object detection of human semen

Malte Stær Nissen^*, Oswin Krause, Kristian Almstrup, Søren Kjærulff, Torben T. Nielsen, Mads Nielsen

^*Corresponding author for this work

Department of Computer Science

5 Citations (Scopus)

Abstract

We compare a set of convolutional neural network (CNN) architectures for the task of segmenting and detecting human sperm cells in an image taken from a semen sample. In contrast to previous work, samples are not stained or washed to allow for full sperm quality analysis, making analysis harder due to clutter. Our results indicate that training on full images is superior to training on patches when class-skew is properly handled. Full image training including up-sampling during training proves to be beneficial in deep CNNs for pixel wise accuracy and detection performance. Predicted sperm cells are found by using connected components on the CNN predictions. We investigate optimization of a threshold parameter on the size of detected components. Our best network achieves 93.87% precision and 91.89% recall on our test dataset after thresholding outperforming a classical image analysis approach.

Original language	English
Title of host publication	Image Analysis : 20th Scandinavian Conference, SCIA 2017, Tromsø, Norway, June 12–14, 2017, Proceedings, Part I
Editors	Puneet Sharma, Filippo Maria Bianchi
Number of pages	10
Volume	Part 1
Publisher	Springer
Publication date	2017
Pages	397-406
ISBN (Print)	978-3-319-59125-4
ISBN (Electronic)	978-3-319-59126-1
DOIs	https://doi.org/10.1007/978-3-319-59126-1_33
Publication status	Published - 2017
Event	20th Scandinavian Conference on Image Analysis - Tromsø, Norway Duration: 12 Jun 2017 → 14 Jun 2017 Conference number: 20

Conference

Conference	20th Scandinavian Conference on Image Analysis
Number	20
Country/Territory	Norway
City	Tromsø
Period	12/06/2017 → 14/06/2017

Series	Lecture notes in computer science
Volume	10269
ISSN	0302-9743

Keywords

Convolutional neural networks
Deep learning
Fertility examination
Human sperm
Segmentation

Access to Document

10.1007/978-3-319-59126-1_33

https://arxiv.org/pdf/1704.00498.pdf

Cite this

Nissen, M. S., Krause, O., Almstrup, K., Kjærulff, S., Nielsen, T. T., & Nielsen, M. (2017). Convolutional neural networks for segmentation and object detection of human semen. In P. Sharma, & F. M. Bianchi (Eds.), Image Analysis: 20th Scandinavian Conference, SCIA 2017, Tromsø, Norway, June 12–14, 2017, Proceedings, Part I (Vol. Part 1, pp. 397-406). Springer. https://doi.org/10.1007/978-3-319-59126-1_33

Convolutional neural networks for segmentation and object detection of human semen. / Nissen, Malte Stær; Krause, Oswin; Almstrup, Kristian et al.
Image Analysis: 20th Scandinavian Conference, SCIA 2017, Tromsø, Norway, June 12–14, 2017, Proceedings, Part I. ed. / Puneet Sharma; Filippo Maria Bianchi. Vol. Part 1 Springer, 2017. p. 397-406 (Lecture notes in computer science, Vol. 10269).

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

Nissen, MS, Krause, O, Almstrup, K, Kjærulff, S, Nielsen, TT & Nielsen, M 2017, Convolutional neural networks for segmentation and object detection of human semen. in P Sharma & FM Bianchi (eds), Image Analysis: 20th Scandinavian Conference, SCIA 2017, Tromsø, Norway, June 12–14, 2017, Proceedings, Part I. vol. Part 1, Springer, Lecture notes in computer science, vol. 10269, pp. 397-406, 20th Scandinavian Conference on Image Analysis, Tromsø, Norway, 12/06/2017. https://doi.org/10.1007/978-3-319-59126-1_33

Nissen MS, Krause O, Almstrup K, Kjærulff S, Nielsen TT, Nielsen M. Convolutional neural networks for segmentation and object detection of human semen. In Sharma P, Bianchi FM, editors, Image Analysis: 20th Scandinavian Conference, SCIA 2017, Tromsø, Norway, June 12–14, 2017, Proceedings, Part I. Vol. Part 1. Springer. 2017. p. 397-406. (Lecture notes in computer science, Vol. 10269). doi: 10.1007/978-3-319-59126-1_33

Nissen, Malte Stær ; Krause, Oswin ; Almstrup, Kristian et al. / Convolutional neural networks for segmentation and object detection of human semen. Image Analysis: 20th Scandinavian Conference, SCIA 2017, Tromsø, Norway, June 12–14, 2017, Proceedings, Part I. editor / Puneet Sharma ; Filippo Maria Bianchi. Vol. Part 1 Springer, 2017. pp. 397-406 (Lecture notes in computer science, Vol. 10269).

@inproceedings{0b46cbffddd74b58847e6e9a89c31b9d,

title = "Convolutional neural networks for segmentation and object detection of human semen",

abstract = "We compare a set of convolutional neural network (CNN) architectures for the task of segmenting and detecting human sperm cells in an image taken from a semen sample. In contrast to previous work, samples are not stained or washed to allow for full sperm quality analysis, making analysis harder due to clutter. Our results indicate that training on full images is superior to training on patches when class-skew is properly handled. Full image training including up-sampling during training proves to be beneficial in deep CNNs for pixel wise accuracy and detection performance. Predicted sperm cells are found by using connected components on the CNN predictions. We investigate optimization of a threshold parameter on the size of detected components. Our best network achieves 93.87% precision and 91.89% recall on our test dataset after thresholding outperforming a classical image analysis approach.",

keywords = "Convolutional neural networks, Deep learning, Fertility examination, Human sperm, Segmentation",

author = "Nissen, {Malte St{\ae}r} and Oswin Krause and Kristian Almstrup and S{\o}ren Kj{\ae}rulff and Nielsen, {Torben T.} and Mads Nielsen",

year = "2017",

doi = "10.1007/978-3-319-59126-1_33",

language = "English",

isbn = "978-3-319-59125-4",

volume = "Part 1",

series = "Lecture notes in computer science",

publisher = "Springer",

pages = "397--406",

editor = "Puneet Sharma and Bianchi, {Filippo Maria}",

booktitle = "Image Analysis",

note = "20th Scandinavian Conference on Image Analysis ; Conference date: 12-06-2017 Through 14-06-2017",

}

TY - GEN

T1 - Convolutional neural networks for segmentation and object detection of human semen

AU - Nissen, Malte Stær

AU - Krause, Oswin

AU - Almstrup, Kristian

AU - Kjærulff, Søren

AU - Nielsen, Torben T.

AU - Nielsen, Mads

N1 - Conference code: 20

PY - 2017

Y1 - 2017

N2 - We compare a set of convolutional neural network (CNN) architectures for the task of segmenting and detecting human sperm cells in an image taken from a semen sample. In contrast to previous work, samples are not stained or washed to allow for full sperm quality analysis, making analysis harder due to clutter. Our results indicate that training on full images is superior to training on patches when class-skew is properly handled. Full image training including up-sampling during training proves to be beneficial in deep CNNs for pixel wise accuracy and detection performance. Predicted sperm cells are found by using connected components on the CNN predictions. We investigate optimization of a threshold parameter on the size of detected components. Our best network achieves 93.87% precision and 91.89% recall on our test dataset after thresholding outperforming a classical image analysis approach.

AB - We compare a set of convolutional neural network (CNN) architectures for the task of segmenting and detecting human sperm cells in an image taken from a semen sample. In contrast to previous work, samples are not stained or washed to allow for full sperm quality analysis, making analysis harder due to clutter. Our results indicate that training on full images is superior to training on patches when class-skew is properly handled. Full image training including up-sampling during training proves to be beneficial in deep CNNs for pixel wise accuracy and detection performance. Predicted sperm cells are found by using connected components on the CNN predictions. We investigate optimization of a threshold parameter on the size of detected components. Our best network achieves 93.87% precision and 91.89% recall on our test dataset after thresholding outperforming a classical image analysis approach.

KW - Convolutional neural networks

KW - Deep learning

KW - Fertility examination

KW - Human sperm

KW - Segmentation

UR - http://www.scopus.com/inward/record.url?scp=85020400552&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-59126-1_33

DO - 10.1007/978-3-319-59126-1_33

M3 - Article in proceedings

AN - SCOPUS:85020400552

SN - 978-3-319-59125-4

VL - Part 1

T3 - Lecture notes in computer science

SP - 397

EP - 406

BT - Image Analysis

A2 - Sharma, Puneet

A2 - Bianchi, Filippo Maria

PB - Springer

T2 - 20th Scandinavian Conference on Image Analysis

Y2 - 12 June 2017 through 14 June 2017

ER -

Convolutional neural networks for segmentation and object detection of human semen

Abstract

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this