Details
| Original language | English |
|---|---|
| Title of host publication | European Conference on Biomedical Optics, ECBO_2019 |
| Publisher | OSA - The Optical Society |
| Number of pages | 4 |
| ISBN (print) | 9781510628397 |
| Publication status | Published - 19 Jul 2019 |
| Event | European Conference on Biomedical Optics, ECBO_2019 - Munich, Netherlands Duration: 23 Jun 2019 → 25 Jun 2019 |
Publication series
| Name | Optics InfoBase Conference Papers |
|---|---|
| Volume | Part F142-ECBO 2019 |
| ISSN (Print) | 2162-2701 |
Abstract
Noise in speckle-prone optical coherence tomography tends to obfuscate important details necessary for medical diagnosis. In this paper, a denoising approach that preserves disease characteristics on retinal optical coherence tomography images in ophthalmology is presented. We propose semantic denoising autoencoders, which combine a convolutional denoising autoencoder with a priorly trained ResNet image classifier as regularizer during training. This promotes the perceptibility of delicate details in the denoised images that are important for diagnosis and filters out only informationless background noise. With our approach, higher peak signal-to-noise ratios with PSNR = 31.0 dB and higher classification performance of F1 = 0.92 can be achieved for denoised images compared to state-of-the-art denoising. It is shown that semantically regularized autoencoders are capable of denoising retinal OCT images without blurring details of diseases.
Keywords
- Computer-aided diagnosis, Image classification, Image enhancement, Machine learning
ASJC Scopus subject areas
- Materials Science(all)
- Electronic, Optical and Magnetic Materials
- Engineering(all)
- Mechanics of Materials
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
European Conference on Biomedical Optics, ECBO_2019. OSA - The Optical Society, 2019. 11078_43 (Optics InfoBase Conference Papers; Vol. Part F142-ECBO 2019).
Research output: Chapter in book/report/conference proceeding › Conference contribution › Research › peer review
}
TY - GEN
T1 - Semantic denoising autoencoders for retinal optical coherence tomography
AU - Laves, Max Heinrich
AU - Ihler, Sontje
AU - Kahrs, Lüder Alexander
AU - Ortmaier, Tobias
N1 - Funding information: This research has received funding from the European Union as being part of the EFRE OPhonLas
PY - 2019/7/19
Y1 - 2019/7/19
N2 - Noise in speckle-prone optical coherence tomography tends to obfuscate important details necessary for medical diagnosis. In this paper, a denoising approach that preserves disease characteristics on retinal optical coherence tomography images in ophthalmology is presented. We propose semantic denoising autoencoders, which combine a convolutional denoising autoencoder with a priorly trained ResNet image classifier as regularizer during training. This promotes the perceptibility of delicate details in the denoised images that are important for diagnosis and filters out only informationless background noise. With our approach, higher peak signal-to-noise ratios with PSNR = 31.0 dB and higher classification performance of F1 = 0.92 can be achieved for denoised images compared to state-of-the-art denoising. It is shown that semantically regularized autoencoders are capable of denoising retinal OCT images without blurring details of diseases.
AB - Noise in speckle-prone optical coherence tomography tends to obfuscate important details necessary for medical diagnosis. In this paper, a denoising approach that preserves disease characteristics on retinal optical coherence tomography images in ophthalmology is presented. We propose semantic denoising autoencoders, which combine a convolutional denoising autoencoder with a priorly trained ResNet image classifier as regularizer during training. This promotes the perceptibility of delicate details in the denoised images that are important for diagnosis and filters out only informationless background noise. With our approach, higher peak signal-to-noise ratios with PSNR = 31.0 dB and higher classification performance of F1 = 0.92 can be achieved for denoised images compared to state-of-the-art denoising. It is shown that semantically regularized autoencoders are capable of denoising retinal OCT images without blurring details of diseases.
KW - Computer-aided diagnosis
KW - Image classification
KW - Image enhancement
KW - Machine learning
UR - http://www.scopus.com/inward/record.url?scp=85084442544&partnerID=8YFLogxK
U2 - 10.1117/12.2526936
DO - 10.1117/12.2526936
M3 - Conference contribution
AN - SCOPUS:85084442544
SN - 9781510628397
T3 - Optics InfoBase Conference Papers
BT - European Conference on Biomedical Optics, ECBO_2019
PB - OSA - The Optical Society
T2 - European Conference on Biomedical Optics, ECBO_2019
Y2 - 23 June 2019 through 25 June 2019
ER -