Related Experiment Video
Updated: Jun 20, 2025
07:11
ARL Spectral Fitting as an Application to Augment Spectral Data via Franck-Condon Lineshape Analysis and Color Analysis
Published on: August 19, 2021
2.4K
Application and performance enhancement of FAIMS spectral data for deep learning analysis using generative
Ruilong Zhang1, Xiaoxia Du1, Hua Li1
1School of Life and Environmental Sciences, GuiLin University of Electronic Technology, GuiLin, 541004, China.
Analytical Biochemistry
|July 21, 2024
Summary
Generative Adversarial Networks (GANs) enhance High-field asymmetric ion mobility spectrometry (FAIMS) analysis of complex mixtures by generating diverse spectral data. This improves deep learning model performance without additional experimental costs.
Area of Science:
- Analytical Chemistry
- Spectroscopy
- Artificial Intelligence
Background:
- High-field asymmetric ion mobility spectrometry (FAIMS) is used for complex mixture analysis.
- Deep learning models for FAIMS often suffer from poor recognition due to limited high-quality, diverse data.
Purpose of the Study:
- To address data limitations in FAIMS deep learning analysis.
- To improve the recognition performance of FAIMS spectral data for complex mixtures.
Main Methods:
- A Generative Adversarial Network (GAN) was employed to simulate and generate realistic, diverse spectral data.
- Real FAIMS spectral data from 15 classes were used to train the GAN.
- Generated data was combined with real data at a 1:4 ratio for deep learning model training (VGG and ResNeXt).
Main Results:
- The GAN effectively expanded the dataset size and increased sample diversity.
- The optimal data ratio of 1:4 (real:generated) significantly improved recognition metrics.
- Accuracy increased by up to 24.19%, precision by 23.71%, recall by 21.08%, and F1-score by 24.50%.
Conclusions:
- GANs offer a cost-effective method for augmenting FAIMS spectral data.
- Data augmentation with GANs substantially enhances deep learning model performance for complex mixture analysis using FAIMS.
