Approximately 40 percent of pharmaceuticals have an origin in natural products, compounds originally derived from plant and animal sources. To this day, physiological activity is keenly investigated in the substances that can be extracted from a diverse range of plants. Moreover, while herbal medicine is an endeavour in its own right, the identification of active compounds in so-called herbal remedies can provide a lead for medicinal chemists. New work in the International Journal of Image Mining describes a way to analyse images of powdered plant products, such as leaves, stems, roots, flowers, fruit, and seeds at the microscopic level for accurate identification and classification.
Bhupendra Fataniya and Tanish Zaveri of Nirma University in Ahmedabad, India, have focused on texture features of three plants – liquorice, rhubarb, and datura (dhatura)- all of which are commonly used in herbal medicine. The team points out that the misidentification of herbal products can lead to serious health problems for patients. Classification is usually carried out by examination of the leaves and other components, but obviously this approach is not possible if the plant has been dried and ground to a powder.
In order to find a solution to this issue, the team has turned to microscopy and a convolutional neural network to allow them to examine the shape and texture of particles in a powdered herbal product. By combining textural features and using a support vector machine, K-nearest neighbour and ensemble classifier the team was able to demonstrate identification from powdered products with 94 percent accuracy. However, they were able to improve on this and achieve an accuracy of 99.8% using a cubic-support vector machine classifier.
Fataniya, B.D. and Zaveri, T. (2021) ‘Microscopic image analysis for herbal plant classification’, Int. J. Image Mining, Vol. 4, No. 1, pp.1–23.
No comments:
Post a Comment