Pharmaceutical hyperspectral reconstruction method based on coded aperture snapshot spectral imaging system

Patent metadata

Field	Value
Publication number	US-12524933-B2
Application number	US-202519245915-A
Country	US
Kind code	B2
Filing date	Jun 23, 2025
Priority date	Mar 24, 2023
Publication date	Jan 13, 2026
Grant date	Jan 13, 2026

Abstract

A pharmaceutical hyperspectral reconstruction method based on a coded aperture snapshot spectral imaging (CASSI) system includes: collecting and processing original pharmaceutical hyperspectral images to obtain augmented pharmaceutical hyperspectral images; performing simulated spatial encoding on the augmented pharmaceutical hyperspectral images to obtain encoded measurement images; performing spectral inverse shift on the encoded measurement images, then performing inverse encoding to obtain inversely encoded three-dimensional hyperspectral images, using the augmented pharmaceutical hyperspectral images as target images, and constructing a training set and a testing set according to the inversely encoded three-dimensional hyperspectral images and the target images; constructing a deep symmetric neural reconstruction network, and training and testing the deep symmetric neural reconstruction network; and deploying a tested deep symmetric neural reconstruction network onto the CASSI system, real-time collecting pharmaceutical measurement images using the snapshot coded imaging system, and performing computational reconstruction on the pharmaceutical measurement images to obtain reconstructed three-dimensional hyperspectral images.

First claim

What is claimed is:

1 . A pharmaceutical hyperspectral reconstruction method based on a coded aperture snapshot spectral imaging system, comprising the following steps:
S 1 , collecting original pharmaceutical hyperspectral images using hyperspectral cameras, preprocessing the original pharmaceutical hyperspectral images to obtain preprocessed pharmaceutical hyperspectral images, and performing data augmentation on the preprocessed pharmaceutical hyperspectral images to obtain augmented pharmaceutical hyperspectral images; S 2 , performing simulated spatial encoding on the augmented pharmaceutical hyperspectral images to obtain encoded measurement images; S 3 , performing spectral inverse shift on the encoded measurement images, followed by performing inverse encoding to obtain inversely encoded three-dimensional hyperspectral images, using the augmented pharmaceutical hyperspectral images as target images, jointly using the inversely encoded three-dimensional hyperspectral images and the target images as a dataset for supervised training, and constructing a training set and a testing set according to the dataset; S 4 , constructing a deep symmetric neural reconstruction network, training the deep symmetric neural reconstruction network using the training set, adjusting a network loss of the deep symmetric neural reconstruction network, performing backpropagation on the deep symmetric neural reconstruction network to update network parameters of the deep symmetric neural reconstruction network, thereby to obtain a trained deep symmetric neural reconstruction network, and testing the trained deep symmetric neural reconstruction network using the testing set to obtain a tested deep symmetric neural reconstruction network; and S 5 , deploying the tested deep symmetric neural reconstruction network onto the coded aperture snapshot spectral imaging system, real-time collecting pharmaceutical measurement images using the coded aperture snapshot spectral imaging system, and performing computational reconstruction on the pharmaceutical measurement images to obtain reconstructed three-dimensional hyperspectral images; and wherein in the step S 1 , the collecting original pharmaceutical hyperspectral images using hyperspectral cameras, preprocessing the original pharmaceutical hyperspectral images to obtain preprocessed pharmaceutical hyperspectral images comprises the following steps: S 11 , presetting a band, a band interval and a hyperspectral image size to obtain a preset band, a preset band interval and a preset hyperspectral image size; S 12 , collecting the original pharmaceutical hyperspectral images using the hyperspectral cameras, selecting and combining, according to the preset band and the preset band interval, a plurality of band images from each of the original pharmaceutical hyperspectral images to obtain combined pharmaceutical hyperspectral images corresponding to the original pharmaceutical hyperspectral images respectively, and cropping, according to the preset hyperspectral image size, the combined pharmaceutical hyperspectral images to obtain cropped pharmaceutical hyperspectral images; S 13 , calculating an average clarity of each of the cropped pharmaceutical hyperspectral images; and S 14 , sorting, according to the average clarity of each of the cropped pharmaceutical hyperspectral images, the cropped pharmaceutical hyperspectral images in descending order, and selecting a top 80% of the cropped pharmaceutical hyperspectral images as the preprocessed pharmaceutical hyperspectral images.

2 . The pharmaceutical hyperspectral reconstruction method based on the coded aperture snapshot spectral imaging system as claimed in claim 1 , wherein in the step S 13 , a formula of the calculating an average clarity of each of the cropped pharmaceutical hyperspectral images is as follows:

D
⁡

(
f
)

=

∑
i

⁢

∑
y

⁢

∑
x

⁢



f
i

⁡

(

x
+
2

,
y

)

-

f
i

⁡

(

x
,
y

)



2

÷
N

,

where D(f) represents an average clarity of a f-th cropped pharmaceutical hyperspectral image, f i (x, y) represents a grayscale value of a pixel (x, y) corresponding to a i-th band image in the f-th cropped pharmaceutical hyperspectral image, f i (x+2, y) represents a grayscale value of a pixel (x+2, y) corresponding to the i-th band image in the f-th cropped pharmaceutical hyperspectral image, N represents a number of bands in the f-th cropped pharmaceutical hyperspectral image, and i=1, 2, . . . , N.

3 . The pharmaceutical hyperspectral reconstruction method based on the coded aperture snapshot spectral imaging system as claimed in claim 2 , wherein in the step S 1 , the performing data augmentation on the preprocessed pharmaceutical hyperspectral images to obtain augmented pharmaceutical hyperspectral images comprises the following steps:
S 15 , rotating two-dimensional images corresponding to channels in the preprocessed pharmaceutical hyperspectral images by a preset rotation angle to obtain rotated hyperspectral images; S 16 , concatenating the preprocessed pharmaceutical hyperspectral images with corresponding rotated hyperspectral images to obtain mixed hyperspectral images; and S 17 , performing a plurality of random crops on the mixed hyperspectral images to obtain a plurality of cropped mixed hyperspectral images, and using the plurality of cropped mixed hyperspectral images as the augmented pharmaceutical hyperspectral images.

4

Assignees

• Univ Hunan

Inventors

• Wang Yaonan
• Chen Zhongwei
• Mao Jianxu
• Zhang Hui
• Liu Caiping
• Su Xuesan
• Miao Hui
• Chen Yurong
• Yin Ating
• Zhao Bingrui

Classifications

• G06T12/20 Primary

  Inverse problem, i.e. transformations from projection space into object space · CPC title

• G06T2210/22

  Cropping · CPC title

• G06T2211/441

  AI-based methods, deep learning or artificial neural networks · CPC title

• G06T9/00

  Image coding (bandwidth or redundancy reduction for static pictures H04N1/41; coding or decoding of static colour picture signals H04N1/64; methods or arrangements for coding, decoding, compressing or decompressing digital video signals H04N19/00) · CPC title

• G06N3/08

  Learning methods · CPC title