Automated Tasmanian devil segmentation and devil facial tumour disease classification

Fatih Veysel Nurçin; Niyazi Şentürk; Elbrus Imanov; Sam Thalmann; Karen Fagg

doi:10.1071/WR22155

RESEARCH ARTICLE (Open Access)

Previous Next Contents Vol 51(1)

Automated Tasmanian devil segmentation and devil facial tumour disease classification

Fatih Veysel Nurçin

^A ^* , Niyazi Şentürk ^A , Elbrus Imanov ^B , Sam Thalmann ^C and Karen Fagg ^C

+ Author Affiliations

- Author Affiliations

^A Department of Biomedical Engineering, Near East University, TRNC Mersin 10, Nicosia 99138, Turkey.

^B Department of Computer Engineering, Near East University, TRNC Mersin 10, Nicosia 99138, Turkey.

^C Save the Tasmanian Devil Program, Department of Natural Resources and Environment, Hobart, Tas. 7001, Australia.

^* Correspondence to: fatihnurcin1@gmail.com

Handling Editor: Guangshun Jiang

Wildlife Research 51, WR22155 https://doi.org/10.1071/WR22155

Submitted: 21 September 2022 Accepted: 11 April 2023 Published: 5 June 2023

Abstract

Context

Artificial intelligence algorithms are beneficial for automating the monitoring of threatened species. Devil facial tumour disease (DFTD) is an endemic disease threatening Australia’s Tasmanian devil. The disease is a cancer that can be transmitted from one devil to another during social interactions. Cameras and trapping techniques have been employed to monitor the spread of the disease in the wild. The use of cameras allows for more frequent monitoring of devils than does trapping, but differentiating wounds from tumours in images is challenging, and this requires time and expertise.

Aim

The purpose of this work is to develop a computer vision system to assist in the monitoring of DFTD spread.

Method

We propose a system that involves image segmentation, feature extraction, and classification steps. U-net architecture, global average pooling layer of pre-trained Resnet-18, and support vector machine (SVM) classifiers were employed for these purposes, respectively. In total, 1250 images of 961 healthy and 289 diseased (DFTD) devils were separated into training, validation, and testing sets.

Results

The proposed algorithm achieved 92.4% classification accuracy for the differentiation of healthy devils from those with DFTD.

Conclusion

The high classification accuracy means that our method can help field workers with monitoring devils.

Implications

The proposed approach will allow for more frequent analysis of devils while reducing the workload of field staff. Ultimately, this automation could be expanded to other species for simultaneous monitoring at shorter intervals to facilitate broadened ecological assessments.

Keywords: classification, conservation, DFTD, ecology, segmentation, support vector machine, threatened species, U-net architecture.

References

Abdi H, Williams LJ (2010) Principal component analysis. Wiley Interdisciplinary Reviews: Computational Statistics 2, 433-459.
| Crossref | Google Scholar |

Bode M, Hawkins C, Rout T, Wintle B (2009) Efficiently locating conservation boundaries: Searching for the Tasmanian devil facial tumour disease front. Biological Conservation 142, 1333-1339.
| Crossref | Google Scholar |

Chollet F (2017) Xception: deep learning with depthwise separable convolutions. In ‘Proceedings of the IEEE conference on computer vision and pattern recognition’. pp. 1251–1258. (Hawaii Convention Center: Honolulu, Hawaii)

Christin S, Hervet É, Lecomte N (2019) Applications for deep learning in ecology. Methods in Ecology and Evolution 10, 1632-1644.
| Crossref | Google Scholar |

Department of Natural Resources and Environment (2023) Threatened species–vertebrates. Tasmanian Government. Available at https://nre.tas.gov.au/conservation/threatened-species-and-communities/lists-of-threatened-species/threatened-species-vertebrates [Accessed 23 February 2023]

Drawert B, Matthew S, Powell M, Rumsey B (2022) Saving the devils is in the details: Tasmanian devil facial tumor disease can be eliminated with interventions. Letters in Biomathematics 9, 121-140 Available at https://lettersinbiomath.journals.publicknowledgeproject.org/index.php/lib/article/view/555.
| Google Scholar |

Escalera S, Pujol O, Radeva P (2010) On the decoding process in ternary error-correcting output codes. IEEE Transactions on Pattern Analysis and Machine Intelligence 32, 120-134.
| Crossref | Google Scholar |

Farquharson KA, McLennan EA, Cheng Y, Alexander L, Fox S, Lee AV, Belov K, Hogg CJ (2022) Restoring faith in conservation action: maintaining wild genetic diversity through the Tasmanian devil insurance program. iScience 25, 104474.
| Crossref | Google Scholar |

Fraik AK, Margres MJ, Epstein B, Barbosa S, Jones M, Hendricks S, Schönfeld B, Stahlke AR, Veillet A, Hamede R, McCallum H, Lopez-Contreras E, Kallinen SJ, Hohenlohe PA, Kelley JL, Storfer A (2020) Disease swamps molecular signatures of genetic-environmental associations to abiotic factors in Tasmanian devil (Sarcophilus harrisii) populations. Evolution 74, 1392-1408.
| Crossref | Google Scholar |

Hawkins CE, Baars C, Hesterman H, Hocking GJ, Jones ME, Lazenby B, Mann D, Mooney N, Pemberton D, Pyecroft S, Restani M, Wiersma J (2006) Emerging disease and population decline of an island endemic, the Tasmanian devil Sarcophilus harrisii. Biological Conservation 131, 307-324.
| Crossref | Google Scholar |

He K, Zhang X, Ren S, Sun J (2016) Deep residual learning for image recognition. In ‘Proceedings of the IEEE conference on computer vision and pattern recognition’. pp. 770–778. (IEEE: Las Vegas, USA)

Iglovikov V, Shvets A (2018) TernausNet: U-net with VGG11 encoder pre-trained on ImageNet for image segmentation. arXiv https://doi.org/10.48550/arXiv.1801.05746.
| Google Scholar |

Kälin U, Lang N, Hug C, Gessler A, Wegner JD (2019) Defoliation estimation of forest trees from ground-level images. Remote Sensing of Environment 223, 143-153.
| Crossref | Google Scholar |

Kingma DP, Ba JL (2015) Adam: a method for stochastic optimization. arXiv https://doi.org/10.48550/arXiv.1412.6980.
| Google Scholar |

Krizhevsky A, Sutskever I, Hinton GE (2017) ImageNet classification with deep convolutional neural networks. Communications of the ACM 60, 84-90.
| Crossref | Google Scholar |

Laborda J, Ryoo S (2021) Feature selection in a credit scoring model. Mathematics 9, 746-23.
| Crossref | Google Scholar |

Loh R, Bergfeld J, Hayes D, O’Hara A, Pyecroft S, Raidal S, Sharpe R (2006) The pathology of devil facial tumor disease (DFTD) in Tasmanian Devils (Sarcophilus harrisii). Veterinary Pathology 43, 890-895.
| Crossref | Google Scholar |

Marburg A, Bigham K (2016) Deep learning for benthic fauna identification. In ‘OCEANS 2016 MTS/IEEE Monterey’. pp. 1–5. (IEEE: Monterey, CA, USA)

Miao Z, Gaynor KM, Wang J, Liu Z, Muellerklein O, Norouzzadeh MS, McInturff A, Bowie RC, Nathan R, Yu SX, Getz WM (2019) Insights and approaches using deep learning to classify wildlife. Scientific Reports 9, 8137.
| Crossref | Google Scholar |

Norouzzadeh MS, Nguyen A, Kosmala M, Swanson A, Palmer MS, Packer C, Clune J (2018) Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning. Proceedings of the National Academy of Sciences of the United States of America 115, E5716-E5725.
| Crossref | Google Scholar |

Novitasari DC, Hendradi R, Caraka RE, Rachmawati Y, Fanani NZ, Syarifudin A, Toharudin T, Chen RC (2020) Detection of COVID-19 chest x-ray using support vector machine and convolutional neural network. Communications in Mathematical Biology and Neuroscience 2020, 1-19.
| Crossref | Google Scholar |

Nurçin FV, Imanov E (2021) Segmentation of overlapping red blood cells for malaria blood smear images by U-net architecture. Journal of Medical Imaging and Health Informatics 11, 2190-2193.
| Crossref | Google Scholar |

Pemberton D, Renouf D (1993) A field-study of communication and social-behavior of the Tasmanian devil at feeding sites. Australian Journal of Zoology 41, 507-526.
| Crossref | Google Scholar |

Pisner DA, Schnyer DM (2020) Support vector machine. In ‘Machine learning’. (Eds A Mechelli, S Vieira) pp. 101–121. (Academic Press)

Ronneberger O, Fischer P, Brox T (2015) U-net: convolutional networks for biomedical image segmentation. In ‘International conference on medical image computing and computer-assisted intervention’. (Eds N Navab, J Hornegger, W Wells, A Frangi) pp. 234–241. (Springer International Publishing: Munich, Germany)

Rout TM, Baker CM, Huxtable S, Wintle BA (2018) Monitoring, imperfect detection, and risk optimization of a Tasmanian devil insurance population. Conservation Biology 32, 267-275.
| Crossref | Google Scholar |

Save the Tasmanian Devil Program (2018) Annual monitoring project: 2018 summary report. Tasmanian Government. Available at https://dpipwe.tas.gov.au/Documents/STDP%20Annual%20Monitoring%20Project%202018%20Summary%20Report.pdf [Accessed 16 February 2021]

Shi C, Liu D, Cui Y, Xie J, Roberts NJ, Jiang G (2020) Amur tiger stripes: Individual identification based on deep convolutional neural network. Integrative Zoology 15, 461-470.
| Crossref | Google Scholar |

Shi C, Xu J, Roberts NJ, Liu D, Jiang G (2023) Individual automatic detection and identification of big cats with the combination of different body parts. Integrative Zoology 18, 157-168.
| Crossref | Google Scholar |

Weinstein BG (2018) A computer vision for animal ecology. Journal of Animal Ecology 87, 533-545.
| Crossref | Google Scholar |

Willi M, Pitman RT, Cardoso AW, Locke C, Swanson A, Boyer A, Veldthuis M, Fortson L (2019) Identifying animal species in camera trap images using deep learning and citizen science. Methods in Ecology and Evolution 10, 80-91.
| Crossref | Google Scholar |

Woods GM, Fox S, Flies AS, Tovar CD, Jones M, Hamede R, Pemberton D, Lyons AB, Bettiol SS (2018) Two decades of the impact of Tasmanian devil facial tumor disease. Integrative and Comparative Biology 58, 1043-1054.
| Crossref | Google Scholar |

Woods GM, Lyons AB, Bettiol SS (2020) A devil of a transmissible cancer. Tropical Medicine and Infectious Disease 5, 50-10.
| Crossref | Google Scholar |

Yang W, Wang K, Zuo W (2012) Fast neighborhood component analysis. Neurocomputing 83, 31-37.
| Crossref | Google Scholar |

Yousif H, Yuan J, Kays R, He Z (2017) Fast human-animal detection from highly cluttered camera-trap images using joint background modeling and deep learning classification. In ‘Proceedings – IEEE international symposium on circuits and systems’. pp. 1–4. (IEEE: Baltimore, MD, USA)