Comparison of logistic regression and neural net modeling for prediction of prostate cancer pathologic stage

Robert W. Veltri, Manisha Chaudhari, M. Craig Miller, Edward C. Poole, Gerard J. O'Dowd, Alan Wayne Partin

Research output: Contribution to journalArticle

Abstract

Background: Prostate cancer (PCa) pathologic staging remains a challenge for the physician using individual pretreatment variables. We have previously reported that UroScore™, a logistic regression (LR)-derived algorithm, can correctly predict organ-confined (OC) disease state with >90% accuracy. This study compares statistical and neural network (NN) approaches to predict PCa stage. Methods: A subset (756 of 817) of radical prostatectomy patients was assessed: 434 with OC disease, 173 with capsular penetration (NOC-CP), and 149 with metastases (NOC-AD) in the training sample. Additionally, an OC + NOC-CP (n = 607) vs NOC-AD (n = 149) two-outcome model was prepared. Validation sets included 120 or 397 cases not used for modeling. Input variables included clinical and several quantitative biopsy pathology variables. The classification accuracies achieved with a NN with an error back-propagation architecture were compared with those of LR statistical modeling. Results: We demonstrated >95% detection of OC PCa in three-outcome models, using both computational approaches. For training patient samples that were equally distributed for the three-outcome models, NNs gave a significantly higher overall classification accuracy than the LR approach (40% vs 96%, respectively). In the two-outcome models using either unequal or equal case distribution, the NNs had only a marginal advantage in classification accuracy over LR. Conclusions: The strength of a mathematics-based disease-outcome model depends on the quality of the input variables, quantity of cases, case sample input distribution, and computational methods of data processing of inputs and outputs. We identified specific advantages for NNs, especially in the prediction of multiple-outcome models, related to the ability to pre- and postprocess inputs and outputs.

Original languageEnglish (US)
Pages (from-to)1828-1834
Number of pages7
JournalClinical Chemistry
Volume48
Issue number10
StatePublished - Oct 1 2002

Fingerprint

Logistics
Prostatic Neoplasms
Logistic Models
Neural networks
Mathematics
Neoplasm Staging
Prostatectomy
Biopsy
Pathology
Neoplasm Metastasis
Computational methods
Physicians
Backpropagation

ASJC Scopus subject areas

  • Clinical Biochemistry

Cite this

Veltri, R. W., Chaudhari, M., Miller, M. C., Poole, E. C., O'Dowd, G. J., & Partin, A. W. (2002). Comparison of logistic regression and neural net modeling for prediction of prostate cancer pathologic stage. Clinical Chemistry, 48(10), 1828-1834.

Comparison of logistic regression and neural net modeling for prediction of prostate cancer pathologic stage. / Veltri, Robert W.; Chaudhari, Manisha; Miller, M. Craig; Poole, Edward C.; O'Dowd, Gerard J.; Partin, Alan Wayne.

In: Clinical Chemistry, Vol. 48, No. 10, 01.10.2002, p. 1828-1834.

Research output: Contribution to journalArticle

Veltri, RW, Chaudhari, M, Miller, MC, Poole, EC, O'Dowd, GJ & Partin, AW 2002, 'Comparison of logistic regression and neural net modeling for prediction of prostate cancer pathologic stage', Clinical Chemistry, vol. 48, no. 10, pp. 1828-1834.
Veltri, Robert W. ; Chaudhari, Manisha ; Miller, M. Craig ; Poole, Edward C. ; O'Dowd, Gerard J. ; Partin, Alan Wayne. / Comparison of logistic regression and neural net modeling for prediction of prostate cancer pathologic stage. In: Clinical Chemistry. 2002 ; Vol. 48, No. 10. pp. 1828-1834.
@article{6a0e4eb7763d452ba50c33d7bf7ad625,
title = "Comparison of logistic regression and neural net modeling for prediction of prostate cancer pathologic stage",
abstract = "Background: Prostate cancer (PCa) pathologic staging remains a challenge for the physician using individual pretreatment variables. We have previously reported that UroScore™, a logistic regression (LR)-derived algorithm, can correctly predict organ-confined (OC) disease state with >90{\%} accuracy. This study compares statistical and neural network (NN) approaches to predict PCa stage. Methods: A subset (756 of 817) of radical prostatectomy patients was assessed: 434 with OC disease, 173 with capsular penetration (NOC-CP), and 149 with metastases (NOC-AD) in the training sample. Additionally, an OC + NOC-CP (n = 607) vs NOC-AD (n = 149) two-outcome model was prepared. Validation sets included 120 or 397 cases not used for modeling. Input variables included clinical and several quantitative biopsy pathology variables. The classification accuracies achieved with a NN with an error back-propagation architecture were compared with those of LR statistical modeling. Results: We demonstrated >95{\%} detection of OC PCa in three-outcome models, using both computational approaches. For training patient samples that were equally distributed for the three-outcome models, NNs gave a significantly higher overall classification accuracy than the LR approach (40{\%} vs 96{\%}, respectively). In the two-outcome models using either unequal or equal case distribution, the NNs had only a marginal advantage in classification accuracy over LR. Conclusions: The strength of a mathematics-based disease-outcome model depends on the quality of the input variables, quantity of cases, case sample input distribution, and computational methods of data processing of inputs and outputs. We identified specific advantages for NNs, especially in the prediction of multiple-outcome models, related to the ability to pre- and postprocess inputs and outputs.",
author = "Veltri, {Robert W.} and Manisha Chaudhari and Miller, {M. Craig} and Poole, {Edward C.} and O'Dowd, {Gerard J.} and Partin, {Alan Wayne}",
year = "2002",
month = "10",
day = "1",
language = "English (US)",
volume = "48",
pages = "1828--1834",
journal = "Clinical Chemistry",
issn = "0009-9147",
publisher = "American Association for Clinical Chemistry Inc.",
number = "10",

}

TY - JOUR

T1 - Comparison of logistic regression and neural net modeling for prediction of prostate cancer pathologic stage

AU - Veltri, Robert W.

AU - Chaudhari, Manisha

AU - Miller, M. Craig

AU - Poole, Edward C.

AU - O'Dowd, Gerard J.

AU - Partin, Alan Wayne

PY - 2002/10/1

Y1 - 2002/10/1

N2 - Background: Prostate cancer (PCa) pathologic staging remains a challenge for the physician using individual pretreatment variables. We have previously reported that UroScore™, a logistic regression (LR)-derived algorithm, can correctly predict organ-confined (OC) disease state with >90% accuracy. This study compares statistical and neural network (NN) approaches to predict PCa stage. Methods: A subset (756 of 817) of radical prostatectomy patients was assessed: 434 with OC disease, 173 with capsular penetration (NOC-CP), and 149 with metastases (NOC-AD) in the training sample. Additionally, an OC + NOC-CP (n = 607) vs NOC-AD (n = 149) two-outcome model was prepared. Validation sets included 120 or 397 cases not used for modeling. Input variables included clinical and several quantitative biopsy pathology variables. The classification accuracies achieved with a NN with an error back-propagation architecture were compared with those of LR statistical modeling. Results: We demonstrated >95% detection of OC PCa in three-outcome models, using both computational approaches. For training patient samples that were equally distributed for the three-outcome models, NNs gave a significantly higher overall classification accuracy than the LR approach (40% vs 96%, respectively). In the two-outcome models using either unequal or equal case distribution, the NNs had only a marginal advantage in classification accuracy over LR. Conclusions: The strength of a mathematics-based disease-outcome model depends on the quality of the input variables, quantity of cases, case sample input distribution, and computational methods of data processing of inputs and outputs. We identified specific advantages for NNs, especially in the prediction of multiple-outcome models, related to the ability to pre- and postprocess inputs and outputs.

AB - Background: Prostate cancer (PCa) pathologic staging remains a challenge for the physician using individual pretreatment variables. We have previously reported that UroScore™, a logistic regression (LR)-derived algorithm, can correctly predict organ-confined (OC) disease state with >90% accuracy. This study compares statistical and neural network (NN) approaches to predict PCa stage. Methods: A subset (756 of 817) of radical prostatectomy patients was assessed: 434 with OC disease, 173 with capsular penetration (NOC-CP), and 149 with metastases (NOC-AD) in the training sample. Additionally, an OC + NOC-CP (n = 607) vs NOC-AD (n = 149) two-outcome model was prepared. Validation sets included 120 or 397 cases not used for modeling. Input variables included clinical and several quantitative biopsy pathology variables. The classification accuracies achieved with a NN with an error back-propagation architecture were compared with those of LR statistical modeling. Results: We demonstrated >95% detection of OC PCa in three-outcome models, using both computational approaches. For training patient samples that were equally distributed for the three-outcome models, NNs gave a significantly higher overall classification accuracy than the LR approach (40% vs 96%, respectively). In the two-outcome models using either unequal or equal case distribution, the NNs had only a marginal advantage in classification accuracy over LR. Conclusions: The strength of a mathematics-based disease-outcome model depends on the quality of the input variables, quantity of cases, case sample input distribution, and computational methods of data processing of inputs and outputs. We identified specific advantages for NNs, especially in the prediction of multiple-outcome models, related to the ability to pre- and postprocess inputs and outputs.

UR - http://www.scopus.com/inward/record.url?scp=0036791345&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0036791345&partnerID=8YFLogxK

M3 - Article

C2 - 12324513

AN - SCOPUS:0036791345

VL - 48

SP - 1828

EP - 1834

JO - Clinical Chemistry

JF - Clinical Chemistry

SN - 0009-9147

IS - 10

ER -