1978 年 17 巻 1 号 p. 1-11
This study was undertaken to develop an automated analytic system for recognition of atypical cells derived from dysplasia and carcinoma in situ of the cervix. Routine, cervical smears stained with Papanicolaou method were used. The cell samples were classified into two different cell patterns, i.e. parabasal dysplastic cells, one with finely granular chromatin and the other with coarsely granular chromatin. These cells were measured by means of Zeiss's SMP-05 connected on line to PDP 12 computer. A measuring spot of 1.0×1.0 micron was used and all measurements were carried out at a wavelength 550 nm. Digitized mirror images of individual cells were first printed out and paper punch tapes of absorbance data from each cell samples were concurrently prepared (paper tape data
I). The analytic system for cell recognition developed by the research group of the Tokyo Metropolitan Detection Center of Cancer was composed of following three systems system of preanalysis for HITAC 10 II (A), system for contour line analysis (B) and system for graph analysis
(C). System A is one which converts absorbance data measured with PDP 12 computer into paper tape data II for HITAC 10 II. Those converted data were introduced into system B. Area of the inside of LOOP, RATIO=length around LOOP (L) squared/area (5), INSUM=total of absorbance value within LOOP, INNO=number of spot within LOOP, MIN=minimum absorbance value within LOOP, MAX=maximum absorbance value within LOOP, XO, YO=barycentric co-ordinates of LOOP and SLOPE=average slope of absorbance value etc. are calculated and printed out as basic data. These basic data were also preserved in paper tape data III as original data prepared by contour line analysis. When these paper tape data were introduced into system C, graphs of AREA, SLOPE and RATIO were made, and selected data are further printed out. These data are easy to understand, because most important parameters of basic data mentioned above are only picked up.
In case of decision and identification of cell pattern, area of the nucleus, width of the nuclear border and number of intranuclear granula were used as important parameters. On the basis of those parameters the indication graph for decision of cell pattern and histologic typing of lesions was drawn. As a result of this investigation the relation between both points was recognized, that there is an intimate relation between indication of cell pattern and histologic diagnosis of lesions.
We intend to develop the study on complete automation of the analysis system, reduction of calculating times and printing out of indication graphs of the relation between cell pattern and histologic diagnosis of lesions.