D. Pantazopoulos et al., BACK-PROPAGATION NEURAL-NETWORK IN THE DISCRIMINATION OF BENIGN FROM MALIGNANT LOWER URINARY-TRACT LESIONS, The Journal of urology, 159(5), 1998, pp. 1619-1623
Purpose: We investigated the potential value of morphometry and artifi
cial intelligence tools to discriminate between benign and malignant l
ower urinary tract lesions. Materials and Methods: The lesions include
d lithiasis in 50 cases, inflammation in 61, benign prostatic hyperpla
sia in 99, carcinoma in situ in 5, and grade I and grades II and III t
ransitional cell carcinoma of the bladder in 71 and 184, respectively.
Images of routine processed voided urine smears stained by the Giemsa
technique were analyzed using a custom image analysis system, providi
ng a data set of 45,452 cells. A neural net model of the back propagat
ion type was used to discriminate benign from malignant cells based on
the extracted morphometric and textural features. Data from 13,636 ra
ndomly selected cells (30% of the total data) were used as a training
set and the data from the remaining 31,816 cells comprised the test se
t. In a similar attempt to discriminate at the patient level data on 3
0% of those randomly selected were used to train a back propagation ne
ural net and data on the remaining 329 were used for testing. Results:
Application of the back propagation neural net enabled the correct cl
assification of 95.34% of benign and 86.71% of malignant cells with ov
erall 90.57% accuracy. At the patient level the back propagation neura
l net enabled the correct classification of 100% of those with benign
and 94.51% of those with malignant disease with overall 96.96% accurac
y. Conclusions: The use of neural nets and image morphometry may incre
ase the speed of cytological diagnosis and the diagnostic accuracy of
voided urine cytology.