The inclusion of nuclear area improves the Paris system for reporting urinary cytology.

OBJECTIVE

The Paris System for Reporting Urinary Cytology (TPS) is a well-known urinary diagnostic model; however, occasional false-positives are a problem. To address this issue, we developed an improved algorithm (IA), based on additional cytological features, for TPS diagnosis.

METHODS

Cytological features were evaluated in 29 hard-to-classify cases, including 22 malignant cases and seven benign cases, using image analysis. The optimal IA was determined using the area under the receiver operating characteristic curve as an index. Re-evaluation was performed by applying measured values to the TPS and IA algorithms.

RESULTS

Using TPS, 12 of the 22 malignant cases were reassigned to a more appropriate category, and the remaining 10 malignant cases remained hard-to-classify. Two of the seven benign cases were classified as suspicious for high-grade urothelial carcinoma, and the remaining five benign cases remained in the original category. The IA, which included nuclear area as a parameter, showed the same diagnostic sensitivity as TPS, and three of the seven benign cases were reassessed as negative. Thus, the positive and negative predictive values of the IA were higher than those of TPS (84.6% and 100% vs 75.9% and 0%).

CONCLUSIONS

The newly developed IA is a practical algorithm with which to address the limitations of TPS and thus may contribute to improved diagnostic accuracy.