Joint Optimization of Convolutional Neural Network and Image Preprocessing Selection for Embryo Grade Prediction in In Vitro Fertilization

Kento Uchida*, Shota Saito, Panca Dewi Pamungkasari, Yusei Kawai, Ita Fauzia Hanoum, Filbert H. Juwono, Shinichi Shirakawa

*Corresponding author for this work

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

Abstract

The convolutional neural network (CNN) is a standard tool for image recognition. To improve the performance of CNNs, it is important to design not only the network architecture but also the preprocessing of the input image. Extracting or enhancing the meaningful features of the input image in the preprocessing stage can help to improve the CNN performance. In this paper, we focus on the use of the well-known image processing filters, such as the edge extraction and denoising, and add the preprocessed images to the input of CNNs. As the optimal filter selection depends on dataset, we develop a joint optimization method of CNN and image processing filter selection. We represent the image processing filter selection by a binary vector and introduce the probability distribution of the vector. To derive the gradient-based optimization algorithm, we compute the gradients of weight and distribution parameters on the expected loss under the distribution. The proposed method is applied to an embryo grading task for in vitro fertilization, where the embryo grade is assigned based on the morphological criterion. The experimental result shows that the proposed method succeeds to reduce the test error by more than 8% compared with the naive CNN models.

Original languageEnglish
Title of host publicationAdvances in Visual Computing - 14th International Symposium on Visual Computing, ISVC 2019, Proceedings
EditorsGeorge Bebis, Bahram Parvin, Richard Boyle, Darko Koracin, Daniela Ushizima, Sek Chai, Shinjiro Sueda, Xin Lin, Aidong Lu, Daniel Thalmann, Chaoli Wang, Panpan Xu
PublisherSpringer
Pages14-24
Number of pages11
ISBN (Print)9783030337223
DOIs
Publication statusPublished - 2019
Externally publishedYes
Event14th International Symposium on Visual Computing, ISVC 2019 - Nevada, United States
Duration: 7 Oct 20199 Oct 2019

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11845 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference14th International Symposium on Visual Computing, ISVC 2019
Country/TerritoryUnited States
CityNevada
Period7/10/199/10/19

Keywords

  • Convolutional neural network
  • Embryo grading
  • Image processing filter
  • In vitro fertilization

Fingerprint

Dive into the research topics of 'Joint Optimization of Convolutional Neural Network and Image Preprocessing Selection for Embryo Grade Prediction in In Vitro Fertilization'. Together they form a unique fingerprint.

Cite this