Identification of the Snail Oncomelania hupensis Lindoensis as Schistotomiasi Host Using CNN

Muh Alif Alghifari (Information Technology Department, Faculty of Engineering, Universitas Tadulako, Jl. Soekarno Hatta No KM 9 Palu 94148, Central Sulawesi, Indonesia)
Hajra Rasmita Ngemba (Information Technology Department, Faculty of Engineering, Universitas Tadulako, Jl. Soekarno Hatta No KM 9 Palu 94148, Central Sulawesi, Indonesia)
Junus Widjaja (National Research and Innovation Agency, Center for Public Health and Nutrition Research, Jl. M.H. Thamrin No. 8, Jakrta Pusat 10340, Jakarta, Indonesia)
Syaiful Hendra (Information Technology Department, Faculty of Engineering, Universitas Tadulako, Jl. Soekarno Hatta No KM 9 Palu 94148, Central Sulawesi, Indonesia)
Muhammad Yazdi (Information Technology Department, Faculty of Engineering, Universitas Tadulako, Jl. Soekarno Hatta No KM 9 Palu 94148, Central Sulawesi, Indonesia)
Yuri Yudhaswana Joefrie (Information Technology Department, Faculty of Engineering, Universitas Tadulako, Jl. Soekarno Hatta No KM 9 Palu 94148, Central Sulawesi, Indonesia)

Advance Sustainable Science, Engineering and Technology (ASSET)
Vol 5, No 3 (2023): August-October

Schistosomiasis is declared by the World Health Organization as a neglected tropical disease. In Indonesia, schistosomiasis is endemic in three regions of Central Sulawesi. In 2022, the schistosomiasis prevalence rate in humans was 1.44%, far from the government's target of 0% prevalence in humans, snails, and mammals by 2025. The role of technology is to identify O.hupensis Lindoensis snails as schistosomiasis hosts among snails in schistosomiasis endemic areas. This system can make it easier for people to recognize O.hupensis Lindoensis snails and can speed up the identification process and reduce survey costs for officers. The identification system is made with digital image processing techniques using the CNN algorithm with Mobile Net architecture. Model updating in the form of 4 classes with 1200 image data. The results of training accuracy of 93% and validation accuracy of 87% were obtained. The training loss function is 0.17, and the validation loss is 0.33