Neural network-based digitized cell image diagnosis of bladder wash cytology

In this pilot study, we tested whether it is possible to apply neural network-based diagnostics on bladder washings to detect urothelial cell carcinoma of the bladder. Eighty-five bladder wash (BW) samples were chosen at random from our own database. Cystoscopy, histology, and follow-up data concerning tumor recurrence were available. Each slide was scanned by the neural network-based digitized cell image system. The neural network-based diagnosis (NNBD) was based on 128 digitized cell images provided by the system. The light microscopic diagnosis (LMD) was rendered by an experienced cytopathologist using the same terminology, i.e., negative, low-grade tumor, and high-grade tumor. Finally, an automatic QUANTICYT analysis was performed on the same material, with as classification low, intermediate, and high risk. The sensitivity for diagnosing a histologically confirmed tumor was for NNBD 92%, for LMD 50%, and for QUANTICYT 69%. For the three methods, receiver operating characteristic (ROC) curves were made for the thresholds low grade/intermediate risk and high grade/high risk. For the prediction of a positive cystoscopy, the highest area under the curve (AUC) was found for NNBD, being 0.71. The AUC for LMD was 0.58. QUANTICYT analysis had the highest AUC value (0.62) for predicting tumor recurrence after a negative cystoscopy, with a lower value for NNBD (0.50). These findings indicate that neural network-based diagnosis of bladder washing samples is highly promising.