Herlev数据集（PAP-SEAR (DTU/HERLEV) 数据库和相关研究）

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%% Feature extraction %%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Matlab feature extraction is implemented with the matlab files found in the same directory as this 'README.txt' file. Explanation of the data and the features extracted, can be found in the report accompanied this cd-rom in the section called 'Papsmear Data'. The file 'matlab_function_list.txt' is a list of the used functions, so future versions of matlab can be compatibility tested by checking to see if any changes is made in these function. The classification results is saved in the text file 'results.txt'. Every line is the results from the feature extraction of a single cell and its segmented picture. Every line, contains the features first and its class last. For the new data pictures the classes is: class 1: normal superficial. class 2: normal intermediate. class 3: normal columnar. class 4: mild dysplastic. class 5: moderate dysplastic. class 6: severe dysplastic. class 7: carcinoma in situ. The order the features appear in a line is the following: Feature name Feature Nr. ----------------------------------------------------- nucleus area 1 Cyto area 2 NC ratio 3 Nucleus brightness 4 Cyto brightness 5 Nucleus shortest diameter 6 Nucleus longest diameter 7 Nucleus elongation 8 Nucleus roundness 9 Cytoplasm shortest diamter 10 Cytoplasm longest diamter 11 Cytoplasm elongation 12 Cytoplasm roundness 13 Nucleus perimeter 14 Cytoplasm perimeter 15 Nucleus pos 16 Nucleus number max 17 Nucleus number min 18 Cytoplasm number max 19 Cytoplasm number min 20 ----------------------------------------------------- Matlab is needed to run the files. Matlab needs the following toolboxes: - image processing toolbox - iofun toolbox - general toolbox The iofun and general toolboxes follow with the standard matlab program. The image toolbox on the other hand normally has to be bought separately. To run the feature extraction, the file 'extract_features.m' should be run from within the matlab environment. Comments are added in the different matlab files, to make the reading of these easier.