Two-Dimensional Plot of the Image Analysis of Ovarian Tumor Peritoneal Cytology Specimens to Distinguish Malignant from Negative Cases

Abstract Objective and Methods: Computer-assisted image analysis (CAIA) was used to analyze Papanicolaou-stained peritoneal cytology specimens from 142 patients with ovarian tumors. A two-dimensional plot was then created with the area of the cell/cell cluster on the x-axis and the maximum value of the connected nuclear area within that cell/cell cluster on the y-axis, after which the patterns were classified. Results: Ovarian cancer positive cases displayed a predominantly vertical pattern (V-type) or a diagonal pattern (D-type) cell/cell cluster plots, whereas the majority of the negative cases showed a small cluster pattern (S-type) or a horizontal pattern (H-type). In suspicious cases, almost equal frequencies of V-, D-, and S-types were observed. We found that V- and D-type patterns in negative cases were attributed to the presence of clusters of mesothelial cells, lymphocytes, and histiocytes. However, the amount of cell clusters in the area where positive cells were distributed (positivity rate) was significantly lower in both negative and suspicious cases than in positive cases. Evaluating the plot pattern and positivity rate together allowed us to reliably distinguish between malignant and negative cases. Conclusion: Although previous CAIA of cytological specimens has fundamentally focused on nuclear area, morphology, and color evaluation, our method of combining cell/cell cluster area and the connected nuclear area in two dimensions reflects a part of the algorithm used by humans in cytological diagnosis and can become a promising analytical method in the near future. This study highlights the importance of actively incorporating diagnostic indicators used by humans into CAIA.