Are nuclear texture features a suitable tool for predicting non-organ-confined prostate cancer?

J Urol. 1999 Jul;162(1):258-62. doi: 10.1097/00005392-199907000-00078.

Abstract

Purpose: We investigated the possibility of determining organ confinement of prostate cancer using multiple nuclear texture features determined by fully automated high resolution image analysis combined with preoperative serum PSA levels.

Materials and methods: The study population consisted of 145 patients (61 organ confined and 84 non-organ-confined cases). Nuclear texture features were determined using single cell preparations of radical prostatectomy specimens. Nuclear texture features were extracted and analyzed by multivariate logistic regression analysis in order to build a classifier for distinguishing between organ confined and non-organ-confined tumors. The classifier was designed in a cell by cell model and tested on a case by case analysis.

Results: The predictive probability of the trained classifier in the cell by cell analysis had a sensitivity of 63%, a specificity of 53%, a positive predictive value of 75% and a negative predictive value of 38% and an area under the ROC curve of 0.58. In the case by case analysis the sensitivity was 70%, the specificity was 54%, positive predictive value 78%, negative predictive value 74%, area under the ROC curve 0.62. When preoperative PSA was included in the algorithm, sensitivity raised to 80%, specificity to 60%, the positive predictive value raised to 79%, the negative predictive value to 52% and the area under the ROC curve to 0.70.

Conclusions: In contrast to former studies using tissue sections, our results suggest that nuclear texture features extracted from single cell preparations cannot be used as a reliable parameter for the determination of organ confinement in prostatic adenocarcinomas.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adenocarcinoma / genetics
  • Adenocarcinoma / pathology*
  • Aged
  • Cell Nucleus / pathology*
  • Humans
  • Male
  • Ploidies
  • Predictive Value of Tests
  • Prostatic Neoplasms / genetics
  • Prostatic Neoplasms / pathology*
  • ROC Curve
  • Sensitivity and Specificity