A Novel Hyperspectral Microscopic Imaging System for Evaluating Fresh Degree of Pork

Visible and NIR microscopic hyperspectrum reconstruction from RGB images with deep convolutional neural networks

We investigate the microscopic hyperspectral reconstruction from RGB images with a deep convolutional neural network (DCNN) in this paper. Based on the microscopic hyperspectral imaging system, a homemade dataset consisted of microscopic hyperspectral and RGB image pairs is constructed. For considering the importance of spectral correlation between neighbor spectral bands in microscopic hyperspectrum reconstruction, the 2D convolution is replaced by 3D convolution in the DCNN framework, and a metric (weight factor) used to evaluate the performance reconstructed hyperspectrum is also introduced into the loss function used in training. The effects of the dimension of convolution kernel and the weight factor in the loss function on the performance of the reconstruction model are studied. The overall results indicate that our model can show better performance than the traditional models applied to reconstruct the hyperspectral images based on DCNN for the public and the homemade microscopic datasets. In addition, we furthermore explore the microscopic hyperspectrum reconstruction from RGB images in infrared region, and the results show that the model proposed in this paper has great potential to expand the reconstructed hyperspectrum wavelength range from the visible to near infrared bands.

Application of Microspectral Imaging in Motor and Sensory Nerve Classification

Objective

It aimed to explore the application of the microscopic hyperspectral technique in motor and sensory nerve classification.

Methods

The self-developed microscopic hyperspectral acquisition system was applied to collect the data of anterior and posterior spinal cord sections of white rabbits. The joint correction algorithm was employed to preprocess the collected data, such as noise reduction. On the basis of pure linear light source index, a new pixel purification algorithm based on cross contrast was proposed to extract more regions of interest, which was used for feature extraction of motor and sensory nerves. Besides, the ML algorithm was employed to classify motor and sensory nerves based on feature extraction results.

Results

The joint correction algorithm was adopted to preprocess the data collected by the microscopic hyperspectral technique, so as to eliminate the influence of the incident light source and the system and improve the classification accuracy. The axon and myelin spectrum curves of the two kinds of nerves in the stained specimens had the same trend, but the values of all kinds of spectrum of sensory nerves were higher than those of motor nerves. However, the myelin sheath spectrum curves of motor nerves in the unstained specimens were greatly different from the curves of sensory nerves. The axon spectrum curves had the same trend, but the axon spectrum values of sensory nerves were higher than those of motor nerves. The ML algorithm had high accuracy and fast speed in motor and sensory nerve classification, and the classification effect of stained specimens was better than that of unstained specimens.

Conclusion

The microscopic hyperspectral technique had high feasibility in sensory and motor nerve classification and was worthy of further research and promotion.