Browsing by Subject "2D NMR spectra"

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • Thumbnail Image
    listelement.badge.dso-type Item , listelement.badge.access-status Open Access ,
    Chemical Structure Elucidation from Nuclear Magnetic Resonance Spectra Using CAM Methods with Neural Networks.
    (Tartu Ülikool, 2025) Kroon, Enriko; Kuhn, Stefan, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
    The given thesis investigates the application of visual explainability methods, specifically Grad-CAM and Grad-CAM++, for the identification of chemical substructures in two-dimensional NMR spectra using convolutional neural networks. A desktop application was developed to integrate these algorithms and analyse spectra from HMBC and HSQC experiments. The analysis revealed that chemical mixtures containing additional spectral components led to inconsistent and unreliable heatmaps. The study concludes that Grad-CAM and Grad-CAM++ combined with simple neural network architectures can highlight the pure compound spectra in most cases with varying accuracy, but are insufficient for reliably identifying fatty acid, indole, or steroid substructures in complex spectral mixtures.