SRCV

SRCV extracts chemical-shift values, integrals, and shift ranges from ^13C and ^1H NMR spectral images using machine learning and computer vision to enable automated NMR data extraction for spectral database construction and structure elucidation.

Key Features:

• Function modules: SRCV is organized into four function modules that extract chemical shifts for ^13C and ^1H, determine integral values, and identify shift ranges from NMR spectral images.
• Machine learning for number recognition: Models are trained on curated datasets for number recognition and use a k-nearest neighbor (kNN) algorithm with k = 4, achieving a reported perfect recognition rate.
• Computer vision: Computer vision methods interpret visual patterns in spectral images to locate and parse numeric annotations corresponding to chemical shifts and integrals.
• Performance: Extraction accuracy has been validated with reported per-image processing times between 11 and 21 seconds.

Scientific Applications:

• Literature data extraction: Extraction of ^13C and ^1H NMR data from published spectral images to support construction of comprehensive spectral databases.
• Integration with structure elucidation systems: Automated import of NMR chemical-shift and integral data into computer-assisted structure elucidation workflows.

Methodology:

Training of machine-learning models on curated number-recognition datasets, application of a k-nearest neighbor (kNN) classifier with k = 4, and use of computer vision to interpret spectral image patterns and extract numeric chemical data.