Computer identification of neoplastic urothelial nuclei from the bladder

OBJECTIVE: To introduce computer-based analysis of Feulgen-stained urinary bladder cell nuclei from voided urine to identify neoplastic urothelial nuc lei. STUDY DESIGN: Nuclei from 23 healthy people and 33 patients with urinary bl adder cancer were analyzed. The nuclei from 9 cancer patients with grade GI (stage Ta), 17 with grade G2 (stages Ta, T1, T1a and T2) and 7 with grade G3 (stages Cis, Ta + Tis, T1 and T3b) were analyzed. Image analysis was car ried out by means of a digital image processing system designed by the auth ors. Features describing nuclei were selected as the first step of the proc edure. Then a multistage classifier was constructed to identify positive an d negative cases. RESULTS: The results of this pilot study of a group of 56 patients yielded a 71% correct classification rate in the control group, while a 66% rate wa s obtained among the cancer patients. The sensitivity of the method was 100 % and the specificity was 77%. CONCLUSION: This approach to the identification of neoplastic urothelial nu clei may be sufficiently well developed to be used successfully both in scr eening highrisk populations and in clinical practice.