The use of infrared spectroscopy and machine learning tools for detection ofMeloidogyneinfestations

• DOI: 10.1111/ppa.13246
• Publication Date: 2020-08-02T01:54:45Z
• Authors: Ernesto San‐Blas; Gabriel Paba; Néstor Cubillán; Edgar Portillo; Ana M. Casassa‐Pa...; César González‐Go...; Mayamarú Guerra
• Keywords: 2. Zero hunger; Artificial intelligence; 0303 health sciences; 03 medical and health sciences; Support vector machine; Plant parasitic nematodes; Fourier transformed infrared spectroscopy – attenuated total reflectance (FTIR‐ATR); Genetic algorithms; Meloidogyne enterolobii

Abstract Plant parasitic nematodes are generally soilborne pathogens that attack plants and cause economic losses in many crops. The infested show nonspecific symptoms or, often, symptomless; therefore, diagnosis is performed by taking soil root tissue samples. Here, we a combination of different infrared spectra analysis machine learning algorithms can be used to detect plant nematode infestations before become visible, using leaves instead roots as We found tomato guava with Meloidogyne enterorlobii produced spectral patterns compared uninfested plants. Using partial from 1,450 900/cm the "fingerprint region", principal component analyses indicated after 5 (tomatoes) or 8 weeks (guava), no visible were positively diagnosed. To improve early detection response, modelling. A support vector (SVM) was obtain more robust, accurate models. SVM model contained 34 vectors, 17 for each level. overall performance >97% total accuracy significantly higher, demonstrating absence chance prediction. best prediction infestation obtained at second fourth tomatoes guavas, respectively, reducing diagnostic time half. combined application these techniques reduces processing field laboratory shows enormous advantages avoiding sampling.