Method for image analysis of predicted cell metastasis
A method for image analysis of predicted cell metastasis is provided. A host performs principle component analysis (PCA) for analysis and conversion of reference images into hyperspectral image information. Then an image capture unit sends input images to the host. The host converts the input images into hyperspectral images according to the hyperspectral image information and gets spectral information of the hyperspectral images. Next the host selects a plurality of wave bands corresponding to esophageal cancer cells and performs feature computation of the spectral information to generate corresponding features images. Then the host performs convolution of the feature images with kernels to get a convolution result. Later the host matches the convolution result with sample spectra of sample images to get a comparison result. Lastly the host determines whether metastasis of esophageal cancer cells occurs according to the comparison result.
1. A method for image analysis of predicted cell metastasis comprising the steps of:
performing principle component analysis (PCA) on a plurality of reference images and converting the reference images into a plurality of hyperspectral image information of a plurality of hyperspectral reference images by a host;
capturing and sending at least one input image into the host by an image capture unit;
converting the input image into a hyperspectral image according to the plurality of hyperspectral image information and getting a plurality of spectral information of the hyperspectral image by the host;
selecting a plurality of wave bands corresponding to esophageal cancer cells according to the plurality of the spectral information and performing feature computation to generate a plurality of corresponding feature images by the host;
performing at least one layer of convolution of the feature images with a plurality of kernels to get a convolution result by the host;
performing matching between the convolution result and a plurality of sample spectra of a plurality of sample images to generate a comparison result by the host; and
determining whether the esophageal cancer cells have spread to other parts of body according to the comparison result by the host.
2. The method as claimed in claim 1 , wherein before the step of converting the input image into a hyperspectral image according to the plurality of hyperspectral image information and getting a plurality of spectral information of the hyperspectral image by the host, the method further includes a step of performing data cleaning of the input image by the host.
3. The method as claimed in claim 1 , wherein the step of performing at least one layer of convolution of the feature images with a plurality of kernels to get a convolution result by the host further includes the steps of:
performing at least one layer of convolution of the plurality of feature images to generate at least one feature map by the host;
performing a pooling operation according to the feature maps to reduce size of the feature maps by the host; and
performing a fully-connected operation according to the feature maps for training the input image and getting weight value to obtain a convolution result by the host.
4. The method as claimed in claim 1 , wherein the hyperspectral image information is a plurality of white light images corresponding to the input image and composed of a plurality of color matching functions (CMF), a correction matrix, and a conversion matrix.
5. The method as claimed in claim 1 , wherein in the step of performing matching between the convolution result and a plurality of sample spectra of a plurality of sample images to generate a comparison result by the host, the host reads the sample spectra of the sample images from a database and then matches the sample spectra with the convolution result.
6. The method as claimed in claim 1 , wherein the sample spectra of the sample images are a plurality of spectra corresponding to the esophageal cancer cells of a plurality of endoscopic images.
7. The method as claimed in claim 1 , wherein the sample spectra of the sample images include at least one sample spectrum of at least one image of metastatic esophageal cancer cells and at least one sample spectrum of at least one image of non-metastatic esophageal cancer cells.
8. The method as claimed in claim 7 , wherein the comparison result is metastasis of the esophageal cancer cells when the convolution result is matched with the sample spectrum of the image of metastatic esophageal cancer cells; the comparison result is non-metastasis of the esophageal cancer cells when the convolution result is matched with the sample spectrum of the image of non-metastatic esophageal cancer cells.
9. The method as claimed in claim 1 , wherein the wave bands include 415-450 nm, 500-560 nm, 590-620 nm, and 660-780 nm.