Cervical Cells' DNA Quantitative Analysis Based On Multi-spectral Imaging Technology

Statistics show that in recent years, new cases' number of cervical cancer in developing countries assumes the trend of escalation and the median age is younger, however early screening and diagnosis of cervical cancer have important significance to reduce the incidence and mortality of the patients. Due to the late start of our country's cervical cancer screening, and the unbalanced economic development levels of different regions, and the differences of the cell pathological doctors'number and level, have made serious obstacle for a large-scale cervical screening in our country. As a result, a cervical cancer screening technology with high degree of automation, high specificity and sensitivity, high possibility for a large-scale promotion has urgent realistic demand.At present, the commonly used methods of cervical screening in our country include cytology, HPV-DNA test, VIA/VILI and DNA quantitative analysis. TBS classification is the most commonly used cytology, but it depends on the cytology doctor to observe the cells smear manually, it needs high intensity work and has low efficiency. Although HPV-DNA test has high specificity and sensitivity, but it costs too much to promote in a large scale. Most likely VIA/VILI is a once-in-a-life screening method in the economically undeveloped areas, its price is low as well as its specificity. DNA quantitative analysis is a new kind of detection method which is developed in recent years, it has many advantages such as high screening speed, high sensitivity, and high degree of automation, but its scan report is not intuitive enough and the man-machine conversation capability is bad.This paper proposes a new method of screening for cervical cancer cells which can use both of the TBS classification and DNA quantitative analysis on a cell smear. These two methods can analyze independently at the same time and don't interface with each other, the analysis results of both can complement each other, it has greatly improved the accuracy of the diagnosis. The new method combines the advantages of two screening method, and overcomes the disadvantages of them, it has good application value and practical significance for the large-scale promotion of our country's cervical cancer screening.The difficulty of combining the TBS classification with DNA quantitative analysis on a cell smear to analysis is the absorbance aliasing under the multi-stained environment. This paper has built a new analysis system for cervical cell screening based on multi-spectral imaging analysis technology on the basis of the DNA cell image analysis system of cervical cancer's early detection which is a high-tech achievement in our laboratory supported by the National Key Technologies Research and Development Program of China during the 10th Five-Year Plan Period. We have set up a multi-spectral imaging hardware platform which is dedicated for the multi-stained cell smear's imaging, and established a method to unmix the aliasing absorbance of multiple materials in the multiple-stained environment; on this basis we have realized the unmixing of the image sensor's response of several monochromatic lights, and built a real-time multi-spectral imaging system based on the multi-channel narrowband filter and color CMOS image sensor. The research in this paper includes:1. The common methods of screening for cervical cancer were reviewed, and a new screening method of screening for cervical cancer cells which combining the TBS classification with DNA quantitative analysis on a cell smear had been put forward. This new method has achieved the organic combination of those two kinds of regular screening method for cervical cancer, and improved the accuracy of diagnosis, and has good application value and practical significance.2. The absorbance unmixing matrix model applied in the multi-stained environment was established based on the absorbance's linear superposition principle and multiple linear regression method. The DNA content of nuclei is measured by the DNA's absorbance unmixed from the aliasing absorbance by using the unmixing model's matrix operations, and provides basis for the cervical cancer screening.3. A set of multi-spectral imaging system used to establish the absorbance unmixing matrix model was created, and a set of technological process to obtain the parameters of the absorbance unmixing matrix model was designed. The accuracy of the matrix model is very important to the accuracy of the absorbance's matrix unmixing operations, because the normalization of the parameters actually is the normalization of the material's molar absorption coefficient in each spectral band, so that this process can at the maximum degree rule out the influence of some factors such as the stability of light source, light condition, the stained degree and so on.4. Due to the requirements of real-time and practicality of the new screening method in the practical application, the multi-spectral image unmixing matrix model was created based on multi-channel narrowband filter and color CMOS image sensor, by analyzing the imaging characteristics of multi-spectral imaging technology, and all kinds of spectral characteristics of the device, as well as the differences between CCD and CMOS image sensor. The matrix model can do the unmixing matrix computing operations with the aliasing image sensor response of multiple incident monochromatic lights, and unmix into single spectral band image of each incident light. Using the matrix model has raised the speed of the new screening method, and has high real-time and practicality.5. A hardware circuit system of the real-time multi-spectral imaging system based on FPGA was designed. Inside the FPGA there were several modules which were designed to achieve the respect functions of capturing the output data of the image sensor, decoding, jointing, interpolation and decoding, multi-spectral unmixing matrix operations, image transmission and so on. These modules are designed by using Verilog HDL language, and can achieve the image data's multichannel parallel and pipeline process. Using hardware algorithm to process the image data improved the efficiency, and met the requirements of the real-time and practicality of new screening method in practical application.6. When the confidence level is 99%, some experiments has verified that there is no significant differences between the actual value and the value calculated by the full spectral band absorbance unmixing matrix model, and there is no significant differences between the three spectral band (490-560-650 nm) absorbance unmixing matrix model and full spectral band absorbance unmixing matrix model, and the accuracy of multi-spectral image unmixing matrix model is high enough. By comparing seven group cases'measured data acquired by DNA quantitative analysis based on multi-spectral imaging technology with the data acquired by DNA quantitative analysis based on single spectral band imaging technology, results show that although the coefficients of variation (mean of DI/standard deviation) of diploid cells and tetraploid cells acquired by DNA quantitative analysis based on multi-spectral imaging technology are bigger than those CVs acquired by DNA quantitative analysis based single spectral band imaging technology, but the differ between the diploid cells' means of DI respectively acquired by these two methods is small enough, as well as the tetraploid cells. Generally speaking, DNA quantitative analysis based on multi-spectral imaging technology has the same accuracy and reliability compared with DNA quantitative analysis based on single spectral band imaging technology.