Comparison of Deep Learning-Based Recognition Techniques for Medical and Biomedical Images

Tomas Majtner*, Esmaeil S. Nadimi

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

Abstract

The recognition and classification of medical and biomedical images typically suffer from the problem of a low number of annotated samples. This comes along with the problem of efficient training of the current deep learning frameworks. Therefore, many researchers opt for various techniques which could substitute the traditional training of convolutional neural networks (CNN) from scratch. In this article, we are comparing multiple of these methods, including transfer learning and using the CNNs as feature extractors. The paper contains results on two datasets with different modalities and three different CNN architectures. We demonstrate the high effectiveness of transfer learning and suggest that, in some cases, it is worth retraining more layers at the end of the network for achieving higher accuracy.
Original languageEnglish
Title of host publicationInternational Conference on Computer Analysis of Images and Patterns : CAIP 2019: Computer Analysis of Images and Patterns
EditorsMario Vento, Gennaro Percannella
PublisherSpringer
Publication date22. Aug 2019
Pages492-504
ISBN (Print)978-3-030-29887-6
ISBN (Electronic)978-3-030-29888-3
DOIs
Publication statusPublished - 22. Aug 2019
EventThe 18th International Conference on Computer Analysis of Images and Patterns -
Duration: 2. Sep 20196. Sep 2019
https://caip2019.unisa.it/

Conference

ConferenceThe 18th International Conference on Computer Analysis of Images and Patterns
Period02/09/201906/09/2019
Internet address
SeriesLecture Notes in Computer Science
Volume11678
ISSN0302-9743

Fingerprint

Neural networks
Network architecture
Deep learning

Keywords

  • Image recognition
  • GoogLeNet
  • VGG-16
  • ResNet-50
  • Transfer learning
  • Polyp detection
  • HEp-2 image classification

Cite this

Majtner, T., & S. Nadimi, E. (2019). Comparison of Deep Learning-Based Recognition Techniques for Medical and Biomedical Images. In M. Vento, & G. Percannella (Eds.), International Conference on Computer Analysis of Images and Patterns: CAIP 2019: Computer Analysis of Images and Patterns (pp. 492-504). Springer. Lecture Notes in Computer Science, Vol.. 11678 https://doi.org/10.1007/978-3-030-29888-3_40
Majtner, Tomas ; S. Nadimi, Esmaeil. / Comparison of Deep Learning-Based Recognition Techniques for Medical and Biomedical Images. International Conference on Computer Analysis of Images and Patterns: CAIP 2019: Computer Analysis of Images and Patterns. editor / Mario Vento ; Gennaro Percannella. Springer, 2019. pp. 492-504 (Lecture Notes in Computer Science, Vol. 11678).
@inproceedings{53c6a26c36a94344bf32a12177e33758,
title = "Comparison of Deep Learning-Based Recognition Techniques for Medical and Biomedical Images",
abstract = "The recognition and classification of medical and biomedical images typically suffer from the problem of a low number of annotated samples. This comes along with the problem of efficient training of the current deep learning frameworks. Therefore, many researchers opt for various techniques which could substitute the traditional training of convolutional neural networks (CNN) from scratch. In this article, we are comparing multiple of these methods, including transfer learning and using the CNNs as feature extractors. The paper contains results on two datasets with different modalities and three different CNN architectures. We demonstrate the high effectiveness of transfer learning and suggest that, in some cases, it is worth retraining more layers at the end of the network for achieving higher accuracy.",
keywords = "Image recognition, GoogLeNet, VGG-16, ResNet-50, Transfer learning, Polyp detection, HEp-2 image classification",
author = "Tomas Majtner and {S. Nadimi}, Esmaeil",
year = "2019",
month = "8",
day = "22",
doi = "10.1007/978-3-030-29888-3_40",
language = "English",
isbn = "978-3-030-29887-6",
pages = "492--504",
editor = "Mario Vento and Gennaro Percannella",
booktitle = "International Conference on Computer Analysis of Images and Patterns",
publisher = "Springer",
address = "Germany",

}

Majtner, T & S. Nadimi, E 2019, Comparison of Deep Learning-Based Recognition Techniques for Medical and Biomedical Images. in M Vento & G Percannella (eds), International Conference on Computer Analysis of Images and Patterns: CAIP 2019: Computer Analysis of Images and Patterns. Springer, Lecture Notes in Computer Science, vol. 11678, pp. 492-504, The 18th International Conference on Computer Analysis of Images and Patterns, 02/09/2019. https://doi.org/10.1007/978-3-030-29888-3_40

Comparison of Deep Learning-Based Recognition Techniques for Medical and Biomedical Images. / Majtner, Tomas; S. Nadimi, Esmaeil.

International Conference on Computer Analysis of Images and Patterns: CAIP 2019: Computer Analysis of Images and Patterns. ed. / Mario Vento; Gennaro Percannella. Springer, 2019. p. 492-504 (Lecture Notes in Computer Science, Vol. 11678).

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

TY - GEN

T1 - Comparison of Deep Learning-Based Recognition Techniques for Medical and Biomedical Images

AU - Majtner, Tomas

AU - S. Nadimi, Esmaeil

PY - 2019/8/22

Y1 - 2019/8/22

N2 - The recognition and classification of medical and biomedical images typically suffer from the problem of a low number of annotated samples. This comes along with the problem of efficient training of the current deep learning frameworks. Therefore, many researchers opt for various techniques which could substitute the traditional training of convolutional neural networks (CNN) from scratch. In this article, we are comparing multiple of these methods, including transfer learning and using the CNNs as feature extractors. The paper contains results on two datasets with different modalities and three different CNN architectures. We demonstrate the high effectiveness of transfer learning and suggest that, in some cases, it is worth retraining more layers at the end of the network for achieving higher accuracy.

AB - The recognition and classification of medical and biomedical images typically suffer from the problem of a low number of annotated samples. This comes along with the problem of efficient training of the current deep learning frameworks. Therefore, many researchers opt for various techniques which could substitute the traditional training of convolutional neural networks (CNN) from scratch. In this article, we are comparing multiple of these methods, including transfer learning and using the CNNs as feature extractors. The paper contains results on two datasets with different modalities and three different CNN architectures. We demonstrate the high effectiveness of transfer learning and suggest that, in some cases, it is worth retraining more layers at the end of the network for achieving higher accuracy.

KW - Image recognition

KW - GoogLeNet

KW - VGG-16

KW - ResNet-50

KW - Transfer learning

KW - Polyp detection

KW - HEp-2 image classification

U2 - 10.1007/978-3-030-29888-3_40

DO - 10.1007/978-3-030-29888-3_40

M3 - Article in proceedings

SN - 978-3-030-29887-6

SP - 492

EP - 504

BT - International Conference on Computer Analysis of Images and Patterns

A2 - Vento, Mario

A2 - Percannella, Gennaro

PB - Springer

ER -

Majtner T, S. Nadimi E. Comparison of Deep Learning-Based Recognition Techniques for Medical and Biomedical Images. In Vento M, Percannella G, editors, International Conference on Computer Analysis of Images and Patterns: CAIP 2019: Computer Analysis of Images and Patterns. Springer. 2019. p. 492-504. (Lecture Notes in Computer Science, Vol. 11678). https://doi.org/10.1007/978-3-030-29888-3_40