Hyperspectral imaging-based cutaneous wound classification using neighbourhood extraction 3D convolutional neural network
-
Mücahit Cihan
and
Murat Ceylan
Abstract
Objectives
Hyperspectral imaging is an emerging imaging modality that beginning to gain attention for medical research and has an important potential in clinical applications. Nowadays, spectral imaging modalities such as multispectral and hyperspectral have proven their ability to provide important information that can help to better characterize the wound. Oxygenation changes in the wounded tissue differ from normal tissue. This causes the spectral characteristics to be different. In this study, it is classified cutaneous wounds with neighbourhood extraction 3-dimensional convolutional neural network method.
Methods
The methodology of hyperspectral imaging performed to obtain the most useful information about the wounded and normal tissue is explained in detail. When the hyperspectral signatures of wounded and normal tissues are compared on the hyperspectral image, it is revealed that there is a relative difference between them. By taking advantage of these differences, cuboids that also consider neighbouring pixels are generated, and a uniquely designed 3-dimensional convolutional neural network model is trained with the cuboids to extract both spatial and spectral information.
Results
The effectiveness of the proposed method was evaluated for different cuboid spatial dimensions and training/testing rates. The best result with 99.69% was achieved when the training/testing rate was 0.9/0.1 and the cuboid spatial dimension was 17. It is observed that the proposed method outperforms the 2-dimensional convolutional neural network method and achieves high accuracy even with much less training data. The obtained results using the neighbourhood extraction 3-dimensional convolutional neural network method show that the proposed method highly classifies the wounded area. In addition, the classification performance and the2computation time of the neighbourhood extraction 3-dimensional convolutional neural network methodology were analyzed and compared with existing 2-dimensional convolutional neural network.
Conclusions
As a clinical diagnostic tool, hyperspectral imaging, with neighbourhood extraction 3-dimensional convolutional neural network, has yielded remarkable results for the classification of wounded and normal tissues. Skin color does not play any role in the success of the proposed method. Since only the reflectance values of the spectral signatures are different for various skin colors. For different ethnic groups, The spectral signatures of wounded tissue and the spectral signatures of normal tissue show similar spectral characteristics among themselves.
Funding source: Scientific and Technological Research Council of Turkey
Award Identifier / Grant number: 122E021
Acknowledgment
The authors express their gratitude to Selcuk University’s expert pediatricians Soylu H and Konak M, for their help.
-
Research funding: This work was supported by the Scientific and Technological Research Council of Turkey, 122E021.
-
Author contributions: Cihan M and Ceylan M conceived and designed the study. Cihan M constructed the hardware and software system. Cihan M performed the data collection and analysis. Cihan M wrote the manuscript. Ceylan M reviewed and edited.
-
Competing interests: Authors state no conflict of interest.
-
Informed consent: Informed consent was obtained from all individuals included in this study.
-
Ethical approval: This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Committee of Selcuk University (2022/125).
References
1. Chang, CI. Hyperspectral imaging: techniques for spectral detection and classification, 1st ed. Baltimore: Springer Science & Business Media; 2003.Search in Google Scholar
2. Cihan, M, Ceylan, M, Soylu, H, Konak, M. Fast evaluation of unhealthy and healthy neonates using hyperspectral features on 700–850 Nm wavelengths, ROI extraction, and 3D-CNN. IRBM 2022;43:362–71. https://doi.org/10.1016/j.irbm.2021.06.009.Search in Google Scholar
3. Dremin, V, Marcinkevics, Z, Zherebtsov, E, Popov, A, Grabovskis, A, Kronberga, H, et al.. Skin complications of diabetes mellitus revealed by polarized hyperspectral imaging and machine learning. IEEE Trans Med Imag 2021;40:1207–16. https://doi.org/10.1109/tmi.2021.3049591.Search in Google Scholar PubMed
4. ul Rehman, A, Qureshi, SA. A review of the medical hyperspectral imaging systems and unmixing algorithms’ in biological tissues. Photodiagnosis Photodyn Ther 2021;33:102165. https://doi.org/10.1016/j.pdpdt.2020.102165.Search in Google Scholar PubMed
5. Edelman, GJ, Aalders, MC. Photogrammetry using visible, infrared, hyperspectral and thermal imaging of crime scenes. Forensic Sci Int 2018;292:181–9. https://doi.org/10.1016/j.forsciint.2018.09.025.Search in Google Scholar PubMed
6. Melit Devassy, B, George, S. Forensic analysis of beverage stains using hyperspectral imaging. Sci Rep 2021;11:1–13. https://doi.org/10.1038/s41598-021-85737-x.Search in Google Scholar PubMed PubMed Central
7. Stuart, MB, McGonigle, AJ, Willmott, JR. Hyperspectral imaging in environmental monitoring: a review of recent developments and technological advances in compact field deployable systems. Sensors 2019;19:3071. https://doi.org/10.3390/s19143071.Search in Google Scholar PubMed PubMed Central
8. Özdoğan, G, Lin, X, Sun, DW. Rapid and noninvasive sensory analyses of food products by hyperspectral imaging: recent application developments. Trends Food Sci Technol 2021;111:151–65. https://doi.org/10.1016/j.tifs.2021.02.044.Search in Google Scholar
9. Wang, C, Liu, B, Liu, L, Zhu, Y, Hou, J, Liu, P, et al.. A review of deep learning used in the hyperspectral image analysis for agriculture. Artif Intell Rev 2021;54:5205–53. https://doi.org/10.1007/s10462-021-10018-y.Search in Google Scholar
10. Goldman, RJ, Salcido, R. More than one way to measure a wound: an overview of tools and techniques. Adv Skin Wound Care 2002;15:236–43. https://doi.org/10.1097/00129334-200209000-00011.Search in Google Scholar PubMed
11. Saiko, G, Lombardi, P, Au, Y, Queen, D, Armstrong, D, Harding, K. Hyperspectral imaging in wound care: a systematic review. Int Wound J 2020;17:1840–56. https://doi.org/10.1111/iwj.13474.Search in Google Scholar PubMed PubMed Central
12. Basiri, A, Nabili, M, Mathews, S, Libin, A, Groah, S, Noordmans, HJ, et al.. Use of a multispectral camera in the characterization of skin wounds. Opt Express 2010;18:3244–57. https://doi.org/10.1364/oe.18.003244.Search in Google Scholar PubMed
13. Jeffcoate, WJ, Clark, DJ, Savic, N, Rodmell, PI, Hinchliffe, RJ, Musgrove, A, et al.. Use of HSI to measure oxygen saturation in the lower limb and its correlation with healing of foot ulcers in diabetes. Diabet Med 2015;32:798–802. https://doi.org/10.1111/dme.12778.Search in Google Scholar PubMed
14. Zhang, S, Gnyawali, S, Huang, J, Ren, W, Gordillo, G, Sen, CK, et al.. Multimodal imaging of cutaneous wound tissue. J Biomed Opt 2015;20:016016. https://doi.org/10.1117/1.jbo.20.1.016016.Search in Google Scholar PubMed PubMed Central
15. Yang, Q, Sun, S, Jeffcoate, WJ, Clark, DJ, Musgove, A, Game, FL, et al.. Investigation of the performance of hyperspectral imaging by principal component analysis in the prediction of healing of diabetic foot ulcers. J Imag 2018;4:144. https://doi.org/10.3390/jimaging4120144.Search in Google Scholar
16. Calin, MA, Parasca, SV, Savastru, D, Manea, D. Hyperspectral imaging in the medical field: present and future. Appl Spectrosc Rev 2014;49:435–47. https://doi.org/10.1080/05704928.2013.838678.Search in Google Scholar
17. Calin, MA, Coman, T, Parasca, SV, Bercaru, N, Savastru, RS, Manea, D. Hyperspectral imaging-based wound analysis using mixture-tuned matched filtering classification method. J Biomed Opt 2015;20:046004. https://doi.org/10.1117/1.jbo.20.4.046004.Search in Google Scholar
18. Wahabzada, M, Besser, M, Khosravani, M, Kuska, MT, Kersting, K, Mahlein, AK, et al.. Monitoring wound healing in a 3D wound model by hyperspectral imaging and efficient clustering. PLoS One 2017;12:e0186425. https://doi.org/10.1371/journal.pone.0186425.Search in Google Scholar PubMed PubMed Central
19. Denstedt, M, Pukstad, BS, Paluchowski, LA, Hernandez-Palacios, JE, Randeberg, LL. Hyperspectral imaging as a diagnostic tool for chronic skin ulcers. In: Proc SPIE 8565, photonic therapeutics and diagnostics IX; 2013.10.1117/12.2001087Search in Google Scholar
20. Roy, SK, Krishna, G, Dubey, SR, Chaudhuri, BB. HybridSN: exploring 3-D–2-D CNN feature hierarchy for hyperspectral image classification. Geosci Rem Sens Lett IEEE 2019;17:277–81. https://doi.org/10.1109/lgrs.2019.2918719.Search in Google Scholar
21. Cihan, M, Ceylan, M. KÇ3B-ESA: hiperspektral görüntü sınıflandırması için yeni 3B evrişimli sinir ağı ve uzaktan algılama uygulaması [NE3D-CNN: a new 3D convolutional neural network for hyperspectral image classification and remote sensing application]. Eur J Sci Technol 2020;2020:65–71.10.31590/ejosat.802890Search in Google Scholar
22. Cihan, M. Hiperspektral görüntüleme yöntemi kullanılarak yenidoğan sağlık durumlarının derin öğrenme metotları ile sınıflandırılması [Classification of health status of neonates with deep learning methods using hyperspectral imaging] [Master thesis]. Konya Technical University, Institute of Graduate Education; 2020.Search in Google Scholar
© 2023 Walter de Gruyter GmbH, Berlin/Boston
Articles in the same Issue
- Frontmatter
- Review
- Effectiveness of FES-supported leg exercise for promotion of paralysed lower limb muscle and bone health—a systematic review
- Research Articles
- Stimulation of spinal cord according to recorded theta hippocampal rhythm during rat move on treadmill
- EEG-based driver states discrimination by noise fraction analysis and novel clustering algorithm
- Active fault tolerant deep brain stimulator for epilepsy using deep neural network
- Stacked machine learning models to classify atrial disorders based on clinical ECG features: a method to predict early atrial fibrillation
- A diagnostic method for cardiomyopathy based on multimodal data
- Hyperspectral imaging enables the differentiation of differentially inflated and perfused pulmonary tissue: a proof-of-concept study in pulmonary lobectomies for intersegmental plane mapping
- Hyperspectral imaging-based cutaneous wound classification using neighbourhood extraction 3D convolutional neural network
- The effects of heating rate and sintering time on the biaxial flexural strength of monolithic zirconia ceramics
Articles in the same Issue
- Frontmatter
- Review
- Effectiveness of FES-supported leg exercise for promotion of paralysed lower limb muscle and bone health—a systematic review
- Research Articles
- Stimulation of spinal cord according to recorded theta hippocampal rhythm during rat move on treadmill
- EEG-based driver states discrimination by noise fraction analysis and novel clustering algorithm
- Active fault tolerant deep brain stimulator for epilepsy using deep neural network
- Stacked machine learning models to classify atrial disorders based on clinical ECG features: a method to predict early atrial fibrillation
- A diagnostic method for cardiomyopathy based on multimodal data
- Hyperspectral imaging enables the differentiation of differentially inflated and perfused pulmonary tissue: a proof-of-concept study in pulmonary lobectomies for intersegmental plane mapping
- Hyperspectral imaging-based cutaneous wound classification using neighbourhood extraction 3D convolutional neural network
- The effects of heating rate and sintering time on the biaxial flexural strength of monolithic zirconia ceramics
