% IMPORTANT: The following is UTF-8 encoded.  This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.

@ARTICLE{Thomas:917490,
      author       = {Thomas, Stefan and Behmann, Jan and Rascher, Uwe and
                      Mahlein, Anne-Katrin},
      title        = {{E}valuation of the benefits of combined reflection and
                      transmission hyperspectral imaging data through disease
                      detection and quantification in plant–pathogen
                      interactions},
      journal      = {Journal of plant diseases and protection},
      volume       = {129},
      number       = {3},
      issn         = {1861-3829},
      address      = {Heidelberg},
      publisher    = {Springer},
      reportid     = {FZJ-2023-00703},
      pages        = {505 - 520},
      year         = {2022},
      abstract     = {Previous studies investigating the performance of
                      transmission and reflection datasets for disease detection
                      showed inconsistent results. Within the studies, the
                      performance of transmission imaging varied significantly for
                      the detection of biotroph and necrotrophy plant pathogens,
                      while reflection imaging showed excellent results in both
                      studies. The current study explores the hypothesis that the
                      disparity between these results might be correlated with the
                      different interactions of the respective pathogens with the
                      host plants and the way light interacts with the plant
                      tissue. Pyrenophora teres f. teres and Puccinia hordei—the
                      causative agents of net blotch and brown rust in
                      barley—have been investigated with focus on early-stage
                      detection and quantification (disease severity) of symptoms.
                      Datasets of hyperspectral imaging time-series measurements
                      were analysed through application of multiple data analysis
                      methods (support vector machines; principal component
                      analysis with following distance classifier; spectral
                      decomposition) in order to compare the performance of both
                      datasets for the detection of disease symptoms. It could be
                      shown that transmittance-based brown rust detection (e.g.
                      $12\%$ disease severity) is outperformed by
                      reflectance-based detection (e.g. $36\%$ disease severity)
                      regardless of the algorithm. However, both the detection and
                      quantification of brown rust through transmittance were more
                      accurate than those of powdery mildew in earlier studies.
                      Transmittance and reflectance performed similar for the
                      detection of net blotch disease during the experiments (~
                      $1\%$ disease severity for reflection and transmission).
                      Each data analysis method outperformed manual rating in
                      terms of disease detection (e.g. $15\%$ disease severity
                      according to manual rating and $36\%$ through support vector
                      machines for rust reflection data). Except for the
                      application of a distance classifier on net blotch
                      transmittance data, it could be shown that pixels, which
                      were classified as symptomatic through the data analysis
                      methods while estimated to represent healthy tissue during
                      manual rating, correlate with areas at the edges of manually
                      detected symptoms. The results of this study support the
                      hypothesis that transmission imaging results are highly
                      correlated with the type of plant–pathogen interaction of
                      the respective pathogens, offering new insights into the
                      nature of transmission-based hyperspectral imaging and its
                      application range.},
      cin          = {IBG-2},
      ddc          = {580},
      cid          = {I:(DE-Juel1)IBG-2-20101118},
      pnm          = {2171 - Biological and environmental resources for
                      sustainable use (POF4-217)},
      pid          = {G:(DE-HGF)POF4-2171},
      typ          = {PUB:(DE-HGF)16},
      UT           = {WOS:000744442300002},
      doi          = {10.1007/s41348-022-00570-2},
      url          = {https://juser.fz-juelich.de/record/917490},
}