A Technological Perspective on Traditional Medicine: Classification of Plant Species with Machine Learning

Abstract

Objective: The aim of this study is to determine the morphological characteristics of any plant; that is, to classify it with the method of image processing and machine learning by defining it with features such as leaf shape, color or odor. Method: In this study, plant images obtained from an open access database called kaggle were used as a source for machine learning. After the image learning process, the leaf images of the plants were classified by the Convolutional Neural Network (CNN) method. To verify that the system was working, 100 images of leaves and flowers were taken for each of two different plants, and the number of statistical data was increased to 700 with the ImageData Generator algorithm. Results: It was concluded that the system identified plants with 97% accuracy. The performance of the machine learning algorithm can also be understood from the confusion matrix. In the method followed in this study, diagonal elements 98 and 79 of the confusion matrix were obtained. This indicates that the method we applied is statistically significant. Conclusion: Thanks to the algorithm used in this study, the identification of plants used in traditional and complementary medicine could be made with an accuracy of 97%. With this algorithm, plants containing harmful chemicals can be identified to the user and their use can be prevented. Transferring the algorithm from the computer system to mobile applications by covering more plant varieties will be a guide for future studies.

Keywords

• Medicinal Plants
• Machine Learning
• Image Processing
• Traditional Medicine

Geleneksel Tıbba Teknolojik Bir Bakış: Bitki Türlerinin Makine Öğrenimi ile Sınıflandırılması

Abstract

Amaç: Bu çalışmanın amacı herhangi bir bitkiyi morfolojik özellikleri; yani yaprak biçimi, rengi ya da kokusu gibi özellikleriyle tanımlayarak, görüntü işleme ve makine öğrenmesi yöntemiyle sınıflandırmaktır. Yöntem: Bu çalışmada kaggle adlı açık erişimli veri tabanından elde edilen bitki görüntüleri makine öğrenimi için kaynak olarak kullanıldı. Görüntü öğrenme işlemi yapıldıktan sonra bitkilerin yaprak görüntüleri Evrişimli Sinir Ağı (CNN) yöntemi ile sınıflandırıldı. Sisteminin çalıştığını doğrulamak için iki farklı bitkinin her biri için 100 adet yaprak ve çiçek görüntüsü alınarak Görüntü Veri Üreteci algoritması ile eldeki istatiksel verinin sayısı 700’e arttırıldı. Bulgular: Sisteminin bitkileri % 97 doğrulukla tanımladığı sonucuna varılmıştır. Makine öğrenimi algoritmasının performansı karışıklık matrisinden de anlaşılabilir. Bu çalışmada izlenen yöntemde karışıklık matrisinin köşegenel elemanları 98 ve 79 elde edilmiştir. Bu da uyguladığımız metodun istatistiksel olarak anlamlı olduğunu belirtmektedir. Sonuç: Bu çalışmada kullanılan algoritma sayesinde geleneksel ve tamamlayıcı tıpta kullanılan bitkilerin kimliklemesi %97 doğrulukla yapılabilmiştir. Bu algoritma ile içeriğinde zararlı kimyasalların olduğu bitkiler kullanıcısına tanımlanabilir ve kullanmaları engellenebilir. Algoritmanın daha fazla bitki çeşidini kapsayarak bilgisayar sisteminden mobil uygulamalara aktarılması ileriki çalışmalar için yol gösterici olacaktır.

Keywords

• Tıbbi Bitkiler,
• Makine Öğrenmesi
• Görüntü İşleme
• Geleneksel Tıp

References

1. 1. Ünal M, Dağdeviren HN. Geleneksel ve Tamamlayıcı Tıp Yöntemleri. Euras J Fam Med 2019;8(1):1-9.
2. 2. Mollahaliloğlu S, et al. The New Period in Traditional and Complementary Medicine. Ankara Med J 2015;15(2):102-105.
3. 3. Öztürk YE, Dömbekçi HA, Ünal SN. Geleneksel Tamamlayıcı ve Alternatif Tıp Kullanımı. Bütünleyici ve Anadolu Tıbbı Dergisi 2020;1(3):23–35.
4. 4. Kigen GK, et al. Current trends of traditional herbal medicine practice in Kenya: a review. African Journal of Pharmacology and Therapeutics 2013;2(1):32-37.
5. 5. Nimri LF, Meqdam MM, Alkofahi A. Antibacterial activity of Jordanian medicinal plants. Pharmaceutical biology 1999; 37(3):196-201.
6. 6. IUCN (International Union for Conservation of Nature). Approaches to Conservation of Medicinal Plants and Traditional Knowledge. A Focus on the Chittagong Hill Tracts. Bangladesh Country Office. 2010, 40 P.
7. 7. Joshi AR, Joshi K. Indigenous knowledge and uses of medicinal plants by local communities of the Kali Gandaki Watershed Area, Nepal. Journal of Ethno pharmacology 2000;73(1-2):175–183.
8. 8. Anselem A. Herbs for healing pax herbals Edo State, Nigeria. 2004.

9. 9. Dold AP, and Cocks ML. The trade in medicinal plants in the Eastern Cape Province, South Africa. South African Journal of Science 2002;98:589–598.
10. 10. Donalek, C., Supervised and unsupervised learning, Astronomy Colloquia, California Institute of Technology, USA, 2011.
11. 11. CNL [Internet, cited: 14.08.2023], Available from: https://saturncloud.io/blog/a-comprehensive-guide-to-convolutional- neural-networks-the-eli5-way/
12. 12. Bektaş J, Bektaş Y, Kangal EE. Integrating a novel SRCRN network for segmentation with representative batch-mode experiments for detecting melanoma. Biomedical Signal Processing and Control 2022;71(B):103218
13. 13. Tan JW, et al. Deep Learning for Plant Species Classification using Leaf Vein Morphometric. IEEE/ACMTransactionson Computational Biology and Bioinformatics 2020;17(1):82-90.
14. 14. Cope JS, et al. The extraction of venation from leaf images by evolved vein classifiers and ant colony algorithms. In: Talon B, et al. editors. In International Conference on Advanced Concepts for Intelligent Vision Systems; 2010; Dec 13-16; Sydney, Aust. Berlin: Springer 2010.
15. 15. Anami BS, Suvarna SN, Govardhan A. A combined color, texture and edge features based approach for identification and classification of Indian medicinal plants. International Journal of Computer Applications 2010;6(12):45-51.
16. 16. Kadir A, et al. Neural network application on foliage plant identification. International Journal of Computer Applications 2011; 29(9):15-22.
17. 17. Larese M, et al. Legume identification by leaf vein images classification. Progress in Pattern Recognition, Image Analysis. Computer Vision, and Applications 2010; 447-454.
18. 18. KGL [Internet, cited: 14.08.2023], Available from: https://www.kaggle.com/code/codefantasy/identifying-plants-and-it-s- medicinal-properties
19. 19. Bohera [Internet, cited: 14.08.2023], Available from: http://dhcrop.bsmrau.net/bohera/
20. 20. Oc, [Internet, cited: 14.08.2023] Available from:https://pza.sanbi.org/ochna-serrulata
21. 21. IDG,[Internet,cited:14.08.2023]Availablefrom:https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/imag e/ImageDataGenerator
22. 22. Kamboj VP. Herbal medicine. Current Science 2010;78(1):35–39.
23. 23. Yasukawa K. Medicinal and Edible Plants as Cancer Preventive Agents. In: Vallisuta O, Olimat SM, editors. Drug Discovery Research in Pharmacognosy. Rijeka: In Tech; 2012. pp: 181-208.
24. 24. Shenouda NS, et al. Phytosterol Pygeum africanum regulates prostate cancer in vitro and in vivo. Endocrine 2007;31(1): 72-81.
25. 25. Paul R, Prasad M, Sah NK. Anticancer biology of Azadirachta indica L (neem): a mini review. Cancer Biol Ther 2011;12(6): 467-476.
26. 26. Ngo SN, Williams DB, Head RJ. Rosemary and cancer prevention: preclinical perspectives. Crit. Rev. Food Sci. Nutr 2011; 51(10):946- 954.
27. 27. Sofowora A, Ogunbodede E, Onayad, A. The role and place of medicinal plants in the strategies for disease prevention. African journal of traditional, complementary, and alternative medicines. AJTCAM 2013;10(5):210–229.
28. 28. Oppong SO, et al. A Novel Computer Vision Model for Medicinal Plant Identification Using Log-Gabor Filters and Deep Learning Algorithms. Comput. Intell Neurosci 2022; Article ID 1189509.
29. 29. Kayhan G, Ergün E. Medicinal and Aromatic Plants Identification Using Machine Learning Methods. BAJECE 2020; 8(1): 81- 87.
30. 30. Zhang Y, Wang Y. Recent trends of machine learning applied to multi-source data of medicinal plants. Journal of Pharmaceutical Analysis 2023; https://doi.org/10.1016/j.jpha.2023.07.012.
31. 31. Caruana R, Niculescu-Mizil A. Data mining in metric space: an empirical analysis of supervised learning performance criteria. Proceedings of the 10th ACM SIGKDD international conference on Knowledge discovery and data mining, Seattle- WA, USA, 2004.
32. 32. WHO. Guidelines on developing consumer information on proper use of traditional, complementary and alternative medicine, Italy, 2004.