Application Of Image Analysis Technology To Monitor The Distribution Of Disease And Make A Diagnosis

Image analysis technology is using computer to process of and treat with the target images. Distill the required information and use of the information to analysis and explain the investigated object. Along with the development of computer technology and imaging technology, the application range of image analysis technology is extended to every field in medicine study. Such as application of remote sensing image to surveillance the distribution of disease and cell classification in biology medicine and intelligent diagnosis of medical images. And exert more and more effect of disease prevention and cure.In the first part of this study we collected the prevalence of snail in Jiangning County in 2000, and extracted the Normalized Difference Vegetation Index (NDVI) and Land surface temperature(LST) from ETM+ image by Vector chart of the snail distribution using Arc View8.1 and ERDAS 8.5 software. The relationship between NDVI , LST and the snail distribution were Investigated using Bivariate correlations and stepwise linear regression.The investigation showed that the range of NDVI in each snail habitats in Jiangning County is -0.446~0.483, the range of LST is 13.725~19.10(band6 high gain, 6H), 13.72~18.799(band6 low gain, 6L).In marshland, the correlation analysis of the NDVI and LST extracted from ETM+ satellite image and snail distribution showed: The area ofsnail-breeding site correlated negatively with the minimum NDVI and LST in the buffer region (P<0.01) , and positively with the maximum NDVI and the difference of extremity NDVI and LST in the buffer region (P<0.05) . The positive rate of live snail basket correlated positively with the minimum NDVI maximum and mean NDVI and the difference of extremity LST in the buffer region (P0.05) . The snail density correlated positively with the maximum and mean NDVI, and the difference of extremity LST in the buffer region (P<0.01) . The stepwise linear regression analysis demonstrated: The area of snail-breeding site (Y1) related to the difference of extremity NDVI(NV) and LST(Thv) derived from the imagery of ETM+, The regression formula showed as Y1=3.44*Nv+0.615* Thv +9.553 (P<0.01, R2=0.804). The positive rate of live snail basket(Y2) related to the mean NDVI(Nmean) and the difference of extremity LST(ThV) in the buffer region of ETM+, The regression formula showed as Y2=179.902*Nmean+7.169*Thv (P<0.01, R2=0.759) . The snail density (Y3) related to the mean NDVI(Nmean) and the difference of extremity LST(ThV) in the buffer region of ETM+, The regression formula showed as Y3=6.3*NDVImean+0.41Thv ( P<0.01 , R2=0.823) .In mountainous regions, the correlation analysis of snail distribution and the NDVI as well as LST extracted from ETM+ satellite image showed: The area of snail-breeding site correlated positively with the difference of extremity NDVI and LST in the buffer region (P<0.05) . The positive rate of live snail basket correlated positively with the minimum NDVI and mean NDVI in the buffer region (P<0.05) . The snail density correlated positively with the maximum NDVI, minimum NDVI and mean NDVI in the buffer region (P<0.05) . The stepwise linear regression analysis demonstrated: Thearea of snail-breeding site (Y4) related to the difference of extremity NDVI(NV) and LST(Thy) derived from the imagery of ETM+, The regression formula showed as Y4=10.655*Nv +2.829*Tlv+7.317 (P<0.05, R2=0.311) . The positive rate of live snail basket(Y2) related to the minimum NDVI(Nmean) in the buffer region of ETM+, The regression formula showed as Y2=88.145*Nmin+11.181 (P<0.05, R2=0.127) . The snail density (Y3) related to the mean NDVI(Nmean), The regression formula showed as Y3=9.65*NDVImean (P<0.01, R2=0.526) .In the second part of the study we collected 112 computed tomography(CT) images of hepatocellular carcinoma (HCC) and 121 CT images of other liver disease from CT department of XiJing hospital from 2001 through 2004. Firstly we used Matlab6.x software to preprocess all the CT images and extracted the feature of each CT images for the training and evaluating database. Then we utilized the database to establishment of accessory diagnosis model based on artificial neural network (ANN) in HCC CT Imaging.The validity of this ANN diagnosis model was calculated according to the output of ANN and practical disease of each CT image. The result showed: The sensitivity of the net was 98 percent. The specificity of the net was 96 percent. The Youden's index of the net was 0.94. The positive likelihood ratio of the net was 28.93. The negatively likelihood ratio of the net was 0.02. The accuracy of the net achieved 97 percent. Far more accurate that the accuracy of enhanced HCC CT images. So the diagnosis effect of the ANN model was very well. We extracted the feature of each CT images once more and used the output of ANN and practical disease of each CT image to evaluate the reliability of this ANN diagnosis model. The result showed: The agreement...