Noise reduction processing method and apparatus for a biological tissue image
Noise reduction processing for measured spectrum data is performed without any information loss due to discrete data characteristics of the measured spectrum data. Optical spectra in one or more cross-sections are measured through use of a signal correlated with a substance distributed in a biological tissue, and a biological tissue image having reduced noise is reconstructed from the spectra.
1. A method of acquiring a biological tissue image, the method comprising the step of:
reconstructing a biological tissue image having reduced noise through use of a plurality of measured spectrum data obtained by measuring respective regions of a biological tissue,
wherein the reduction of the noise is performed through use of a technique for machine learning utilizing reference data for the measured spectrum data,
wherein the reference data is generated through utilization of training data, and
wherein the reduction of the noise is performed by generating a classifier through utilization of the training data, dividing an entire spectrum into typical specific spectra through use of the classifier, and reconstructing an image from the typical specific spectra derived from the biological tissue.
2. The method according to claim 1 , wherein the measured spectrum data is one of data of optical spectrum in a range of ultraviolet, visible or infrared light, Raman spectrum and mass spectrum.
3. An apparatus for reconstructing a biological tissue image comprising a central processing unit and memory, said central processing unit and memory cooperating to execute the method of claim 1 .
4. The apparatus according to claim 3 , further comprising a detector for detecting a signal from a sample on a substrate, and an image displaying member for displaying the reconstructed image.
5. The apparatus according to claim 4 , wherein the detector obtains a two-dimensional mass spectrum.
6. The method according to claim 1 , wherein the technique for machine learning is selected from the group consisting of the Fisher's linear discriminant method, a Support Vector Machine (SVM), a decision tree, and a random forest method.
7. The method according to claim 1 , further comprising a step where components other than the typical specific spectra are expressed by 0 with respect to the measured spectrum data.
8. The method according to claim 2 , wherein the measured spectrum data is a two-dimensional mass spectrum which utilizes a mass number (m/z) and a strength data corresponding to the peak of the mass spectrum.