Search Results
A Machine Learning-Based Raman Spectroscopic Assay for the Identification of Burkholderia mallei and Related Species
Burkholderia (B.) mallei, the causative agent of glanders, and B. pseudomallei, the causative agent of melioidosis in humans and animals, are genetically closely related. The high infectious potential of both organisms, their serological cross-reactivity, and similar clinical symptoms in human and animals make the differentiation from each other and other Burkholderia species challenging. The increased resistance against many antibiotics implies the need for fast and robust identification methods. The use of Raman microspectroscopy in microbial diagnostic has the potential for rapid and reliable identification. Single bacterial cells are directly probed and a broad range of phenotypic information is recorded, which is subsequently analyzed by machine learning methods. Burkholderia were handled under biosafety level 1 (BSL 1) conditions after heat inactivation. The clusters of the spectral phenotypes and the diagnostic relevance of the Burkholderia spp. were considered for an advanced hierarchical machine learning approach. The strain panel for training involved 12 B. mallei, 13 B. pseudomallei and 11 other Burkholderia spp. type strains. The combination of top- and sub-level classifier identified the mallei-complex with high sensitivities (>95%). The reliable identification of unknown B. mallei and B. pseudomallei strains highlighted the robustness of the machine learning-based Raman spectroscopic assay. © 2019 by the authors
Computational tissue staining of non-linear multimodal imaging using supervised and unsupervised deep learning
Hematoxylin and Eosin (H&E) staining is the 'gold-standard' method in histopathology. However, standard H&E staining of high-quality tissue sections requires long sample preparation times including sample embedding, which restricts its application for 'real-time' disease diagnosis. Due to this reason, a label-free alternative technique like non-linear multimodal (NLM) imaging, which is the combination of three non-linear optical modalities including coherent anti-Stokes Raman scattering, two-photon excitation fluorescence and second-harmonic generation, is proposed in this work. To correlate the information of the NLM images with H&E images, this work proposes computational staining of NLM images using deep learning models in a supervised and an unsupervised approach. In the supervised and the unsupervised approach, conditional generative adversarial networks (CGANs) and cycle conditional generative adversarial networks (cycle CGANs) are used, respectively. Both CGAN and cycle CGAN models generate pseudo H&E images, which are quantitatively analyzed based on mean squared error, structure similarity index and color shading similarity index. The mean of the three metrics calculated for the computationally generated H&E images indicate significant performance. Thus, utilizing CGAN and cycle CGAN models for computational staining is beneficial for diagnostic applications without performing a laboratory-based staining procedure. To the author's best knowledge, it is the first time that NLM images are computationally stained to H&E images using GANs in an unsupervised manner.
A Review on Data Fusion of Multidimensional Medical and Biomedical Data
Data fusion aims to provide a more accurate description of a sample than any one source of data alone. At the same time, data fusion minimizes the uncertainty of the results by combining data from multiple sources. Both aim to improve the characterization of samples and might improve clinical diagnosis and prognosis. In this paper, we present an overview of the advances achieved over the last decades in data fusion approaches in the context of the medical and biomedical fields. We collected approaches for interpreting multiple sources of data in different combinations: image to image, image to biomarker, spectra to image, spectra to spectra, spectra to biomarker, and others. We found that the most prevalent combination is the image-to-image fusion and that most data fusion approaches were applied together with deep learning or machine learning methods.
Modified PCA and PLS: Towards a better classification in Raman spectroscopy–based biological applications
Raman spectra of biological samples often exhibit variations originating from changes of spectrometers, measurement conditions, and cultivation conditions. Such unwanted variations make a classification extremely challenging, especially if they are more significant compared with the differences between groups to be separated. A classifier is prone to such unwanted variations (ie, intragroup variations) and can fail to learn the patterns that can help separate different groups (ie, intergroup differences). This often leads to a poor generalization performance and a degraded transferability of the trained model. A natural solution is to separate the intragroup variations from the intergroup differences and build the classifier based on merely the latter information, for example, by a well-designed feature extraction. This forms the idea of this contribution. Herein, we modified two commonly applied feature extraction approaches, principal component analysis (PCA) and partial least squares (PLS), in order to extract merely the features representing the intergroup differences. Both of the methods were verified with two Raman spectral datasets measured from bacterial cultures and colon tissues of mice, respectively. In comparison to ordinary PCA and PLS, the modified PCA was able to improve the prediction on the testing data that bears significant difference to the training data, while the modified PLS could help avoid overfitting and lead to a more stable classification. © 2019 The Authors. Journal of Chemometrics published by John Wiley & Sons Ltd