We have developed to cancer screening algorithm of the cell cycle, evaluated by the flow cytometry. Our previous research using our ‘the domain method’ algorithm showed a sensitivity of 98% and specificity of 82% to screen cancer cells. The present study aimed to validate our previous findings by using a linear discriminant analysis (LDA) on other tissues collected from 70 patients with colorectal cancer (60 cancer tissues; 58 normal tissues). The results showed about a 7% decrease in sensitivity from our previous research, thus indicating a need to improve the diagnostic algorithm.
In the present study, aiming to improve the algorithm, called as the domain method, we focused and applied following new features of the cell cycle findings: The second largest peak (PK2) to largest peak (PK1) ratio (PK2／PK1), the G0／G1 and Pseudo domain cell counts, and half-width of the largest peak. These features and our previous features (i.e., debris, S, G2／M, Over G2／M) of the cell cycle were evaluated by the Kolmogorov-Smirnov test. Features for which distributions were judged to discriminate cancer cells from normal cells were adopted. Our findings demonstrated that the adopted features are the PK2／PK1, debris, S, G2／M, and over G2／M domain cell counts.
The statistical analysis was subsequently conducted after screening based on the original histogram largest peak address and presence or absence of a second peak in which PK2／PK1 exceeded 1／2. As the result, the new algorithm showed a sensitivity of >85% and specificity of >95%.
