FN 3-5 | Automatic segmentation and classification of fungal infected tissue using Deep learning

When the human immune system is weakened, fungal infections can lead to a life-threatening condition in humans. In order to determine the type of fungus, pathologists perform tissue biopsies for further examination. A common method of highlighting the fungi is to stain the entire sample slice. An alternative and relatively new method is MALDI-imaging (matrix-assisted laser desorption ionization), in which the mass spectrum of the tissue is measured without destroying the sample. Regardless of which images are used for the pathologist's examination, a lot of time is needed for the examination of the entire tissue due to the big volume image data. Objective: We here use several Deep Learning techniques to distinguish between normal tissue and fungal-infected tissue areas independently of conditions such as the type of organ tissue and fungal genus. We also aim to identify certain features from molecular components in the mass spectra data to their corresponding fungal genus. Materials & Methods: Recently, deep learning has become increasingly important in the field of image-based systems biology. Here we train a convolutional neuronal network (CNN) to segment histopathological tissue slices in order to use these as labels for a second CNN to identify the fungal genus in fungal-infected tissue based on the MALDI images. Results: Preliminary results reveal that the CNN is capable of classifying the fungal genus, for example, Aspergillus fumigatus, Mucorales and Candida albicans. We provide a framework that is able to examine the big volume image data and automatically segment and classify various types of fungi on various types of organs. Conclusion: Deep learning offers a new approach towards the combined analysis of optical and MALDI images capable of performing correlative microscopy-spectroscopy for future online scanning.