The Artificial Neural Network utilization is limited to its complexity. Aim of our study is to organize a simple ANN for a practical utilization.
Methods
We retrospectively reviewed from our database male patients (both outpatients and hospitalized patients) who underwent PSA and free PSA assay. Inclusion criteria was as follows: male older than 45, without a history of prostate cancer. Patients who rejected a proposed prostatic biopsy were excluded. 520 men were included in the study. There were 431 (83%) controls and 89 (17%) had prostate cancer. An ANN was constructed on the basis of data on PSA, fPSA, percentage of free PSA and age.
Results
Of the samples 200 random cases (38.5%) were used to train the system, 100 random cases (19.2%) were used for the test phase and the remains 220 (42.3%) were used for validation. At the ROC analysis the higher area under the curve was for ANN with a significant difference with that of PSA (p=0.022).
Conclusions
We developed a diagnostic algorithm based on both serum data (tPSA, fPSA and %fPSA) and clinical data (age) to enhance the performance of tPSA to discriminate prostate cancer patients. The predictive accuracy of the ANN was superior to that of tPSA. Limiting input neurons, the complexity is reduced so this ANN can be used for daily urological practice.
YoungJ. (Ed). Canver incidence in five continents.Lyon, France, International Agency for Research on Cancer, vol VII1997.
2.
CatalonaW.J., SmithD.S., RatliffT.L.Measurement of prostate-specific antigen in serum as a screening test for prostate cancer. N Engl J Med1991; 324: 1156–61.
3.
WoolfS.H.: screening for prostate cancer with prostate-specific antigen. An examination of evidence. N Engl J Med1995; 1333: 1401–5.
4.
BensonM.C., WhangIs, PantuckA.Prostate specific antigen density: a means of distinguishing benign prostatic hypertrophy and prostate cancer. J Urol1992; 147: 815–6.
5.
CarterH.B., PearsonJd, MetterE.J.Longitudinal evaluation of prostate specific antigen levels in men with and without prostate disease. JAMA1992; 267: 2215–20.
6.
StenmanU-H, LeinonenJ., AlfthanH.A complex between prostate –specific antigen and alpha 1-antichymotrypsin is the major form of prostate-specific antigen in serum of patients with prostatic cancer. Assay of the complex improves clinical sensitivity for cancer. Cancer Res1991; 51: 222–6.
7.
LiljaH., ChristenssonA., DahlenU.Prostate-specific antigen in serum occurs predominandy in complex with alpha 1-antichymotrypsin. Clin Chem1991; 37: 1618–25.
8.
CatalonaW.J., SmithD.S., WolfertR.L.Evaluation of percentage of free serum prostate-specific antigen to improve specificity of prostate cancer screening. JAMA1995; 274: 1214–20.
9.
MarleyG.M., MillerM.C., KattanM.W.Free and complexcd prostate-specific antigen serum ratios to predict probability of primary prostate cancer and benign prostatic hyperplasia. Urology1996; 48: 16–22.
10.
OptenbergS.A., ClarkJ.Y., BrawerM.K.Development of a decision-making tool to predict risk of prostate cancer: the Cancer of the Prostate Risk Index (CAPRI) test. Urology1997; 50: 665–72.
11.
CarlsonG.D., CalvaneseC.B., and PartinA.W.An algorithm combining age, total prostate-specific antigen (PSA), and percent free PSA to predict prostate cancer: results on 4298 cases. Urology1998; 52: 455–61.
12.
KranseR., BeemsterboerP., RietbergenjPredictors for biopsy outcome in the European Randomized Study of Screcning for Prostate Cancer (Rotterdam region). Prostate39: 316–22, 1999.
13.
WeiJ.T., ZhangZ., BarnhillS.D.Understanding artificial neural networks and exploring their potential applications for the practicing urologist. Urology1998; 52: 161–72.
14.
SnowP.B., SmithDS and CatalonaWJ., Artificial neural netwoprks in the diagnosis and prognosis of prostate cancer: a pilot study. J Urol1994; 152: 1923–26.
15.
CrossS.S., HarrisonR.F., KennedyR.L.Introduction to neural networks. Lancet1995; 346: 1075–9.
16.
GuerriereM.R., DetskyA.S.Neural networks: what are they?Ann Intern Med1991; 11: 906–7.
17.
KhannaT.Foundations of neural networks.1th ed.Addison-Wesley Publishing Company, Inc.NY, 1990.
18.
HanleyJ.A., and McNeilB.J.A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology1983; 148: 839.
19.
HanleyJ.A., McNeilB.J.The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology1982, 143: 29.
20.
CatalonaWL, SmithD.S., RatliffT.L., BaslerJ.W.Detection of organ-confined prostate cancer is increased through prostate-specific antigen-based screeening. JAMA1993; 270: 948–54.
21.
CatalonaWJ, RichieP., AhmannF.R., HudsonMA, ScardinoPT, FlaninganR.C., deKernionJ.B., RatliffT.I., KavoussiL.R., DalkinB.L., WatersW.B., MacFarlaneM.T., SouthwickP.C.Comparison of digital rectal examination and serum prostate specific antigen in the carly detectìon of prostate cancer: results of a multicenter clinical trial of 6,630 men. J Urol1994; 151: 1283–90.
22.
ReissiglA., KlockerH., PointnerJ.Usefulness of the ratio free/total prostate-specific antigen in addition to total PSA levels in prostate cancer screening. Urology1996; 48: 62–6.
23.
HorningerW, ReissiglA.Improvement of specificity in PSA based screening by using PSA-transition zone density and percent free PSA in addition to total PSA levels. Prostate1998; 37: 133–9.
24.
HorningerW., BartschG., SnowP.B., BrandtJ.M., PartinA.W.The problem of cutoff levels in a screened population, Cancer2001; 91 (S8): S1667–71.
25.
FinneP., FinneR., AuvinenA., JuuselaH., AroJ., MaattanenL., HakamaM., RannikkoS., TammelaT., StenmanU.Predicting the outcome of prostate biopsy in screen-positive men by a multilayer perceptron network. Urology2000; 56: 418–22.
26.
SargentD.J.Comparison of artificial neural networks with other statistical approaches. Cancer2001; 91: 1636–42.
27.
VirtanenA., GomariM., KranseR., StenmanU.H.Estimation of prostate cancer probability by logistic regression: free and total prostate-specific antigen, digital rectal examination, and heredity are significant. Clin Chem1999; 45: 987–94.
28.
DazziD., TaddeiF., GavariniA., UggeriE., NegroR., PezzarossaA.The control of blood glucose in the critical diabetic patient. A neuro-fuzzy method. J Diabetes Complications2001; 15: 80–7.
29.
CarstenS., CammannH., SemjonowA.Multiventer evaluation of an artificial neural network to increase the prostate cancer detection rate and reduce unnecessary biopsies. Clin Chem2002; 48: 1279–87.
