Hyperspectral Imaging Classification of Fungal Species Using Machine Learning

Hyperspectral Imaging (HSI) offers fast, non-destructive spectral data acquisition, overcoming the limitations of traditional methods. With the HSI technique we aim to improve the identification of fungal species, assisting in the timely diagnosis of fungal infections and automating unhealthy manual processes of microbiological products. This contribution centers on the differentiation of four distinct fungal species by leveraging HSI alongside supervised machine learning techniques, where we conduct comparisons between different classification approaches. After acquiring HSI captures of fungal samples in isolated colonies for each species, they were subjected to an appropriate pre-processing process, which included calibration, correction and application of signal filters. The highest accuracy of 97.12% was achieved by applying the multilayer perceptron (MLP). The results highlight the potential of machine learning methodology applied to HSI for clinical and microbiological environments as an alternative to time-consuming techniques in the work of differentiating fungal species.