Comparing neural networks in the discrimination of benign from malignant lower urinary tract lesions

• Published in: British Journal of Urology 1998-04, Vol.81 (4), p.574-579
• Main Authors: PANTAZOPOULOS, D, KARAKITSOS, P, IOKIM-LIOSSI, A, POULIAKIS, A, DIMOPOULOS, K
• Format: Article
• Language: English
• Subjects: back propagation; Biological and medical sciences; Carcinoma in Situ - diagnosis; Carcinoma, Transitional Cell - diagnosis; Computerized, statistical medical data processing and models in biomedicine; Cystitis - diagnosis; Diagnosis, Differential; Humans; learning vector quantizer; Male; Medical sciences; Models and simulation; morphometry; Nephrology. Urinary tract diseases; Neural networks; Neural Networks (Computer); Prostatic Hyperplasia - diagnosis; Sensitivity and Specificity; transitional cell carcinoma of the bladder; Tumors of the urinary system; Urinary Bladder Calculi - diagnosis; Urinary Bladder Neoplasms - diagnosis; Urinary tract. Prostate gland
• ISSN: 0007-1331; 1464-410X
• DOI: 10.1046/j.1464-410x.1998.00587.x

Objective  To compare the performance of two different neural networks (NNs) in the discrimination of benign and malignant lower urinary tract lesions. Materials and methods  A group of patients was evaluated, comprising 50 cases of lithiasis, 61 of inflammation, 99 of benign prostatic hyperplasia (BPH), five of insitu carcinoma, 71 of grade I transitional cell carcinoma of the bladder (TCCB), and 184 of grade II and grade III TCCB. Images of routinely processed voided urine smears were stained using the Giemsa technique and analysed using an image‐analysis system, providing a dataset of 45 452 cells. Two NN models of the back propagation (BP) and learning vector quantizer (LVQ) type were used to discriminate benign from malignant cells and lesions, based on morphometric and textural features. The data from 13 636 randomly selected cells (30% of the total data) were used as a training set and data from the remaining 31 816 cells comprised the test set. Similarly, in an attempt to discriminate patients, 30% of the cases, selected randomly, were used to train a BP and an LVQ NN, with the remaining 329 cases used for the test set. The data used for training and testing were the same for the two kinds of classifiers. Results  The two NNs gave similar results, with an overall accuracy of discrimination of ≈90.5% at the cellular level and of ≈97% for individual patients. There were no statistically significant differences between the two NNs at the cellular or patient level. Conclusions  The use of NNs and image morphometry could increase the diagnostic accuracy of voided urine cytology; despite the different nature of the two classifiers, the results obtained were very similar.