Intelligent Diagnosis of Ganoderma boninense Infection in Oil Palm via Integration of Raman Spectroscopy and High-Accuracy Classification Models

Ganoderma boninense is identified as the primary fungal pathogen that causes basal stem rot disease in the oil palm industry, and this infestation is typically only detectable at the late stage of infection. Conventional methods and modern molecular techniques such as polymerase chain reaction (PCR) are found to have limitations in terms of long analysis time allocation and the issue of full dependence on in vitro sample processing. To identify G. boninense fungal samples via the Raman spectroscopy method, a 532 nm emission laser from a confocal system was utilized to analyze two types of fungal cultures (monokaryon and dikaryon). Using the same laser configuration, a total of 36 infected leaf tissue samples and 6 healthy controls were systematically scanned based on their pathological status, which were selected from a total of 70 infected samples and 10 control samples prepared for the signal processing phase. The spectrum pre-processing phase and machine learning algorithms were then executed on the leaf samples. The results of the study indicate that the Gaussian Naive Bayes classification model integrated with 5-fold cross-validation shows high classification capability in classifying the differences between healthy and G. boninense-infected leaves. This optimum performance was achieved using four specific predictors: three Full Width at Half Maximum (FWHM) values derived from phenolic compounds (P1), chlorophyll (P2), and phenylalanine (P3), as well as one peak band intensity feature from P3. To overcome the possibility of overfitting issues due to the relatively limited sample size, future studies will involve a comprehensive increase in sample size. In conclusion, the results of this study prove that Raman spectroscopy offers a high-performance and non-invasive methodology for early detection as well as accurate classification of oil palm health status.