Related Experiment Videos
Computer-derived nuclear "grade" and breast cancer prognosis
W H Wolberg1, W N Street, D M Heisey
1Department of Surgery, University of Wisconsin, Madison, USA.
Analytical and Quantitative Cytology and Histology
|August 1, 1995
Summary
Computer analysis of nuclear features from fine needle aspiration (FNA) samples offers a more accurate prognosis for invasive breast cancer than traditional methods. These objective measurements surpass tumor size and lymph node status in predicting patient outcomes.
Area of Science:
- Computational pathology
- Oncology
- Medical imaging analysis
Background:
- Visual assessment of nuclear grade is subjective but crucial for breast cancer prognosis.
- Objective, quantitative analysis of nuclear features is needed to improve prognostic accuracy.
Purpose of the Study:
- To evaluate the prognostic significance of computer-derived nuclear features in invasive breast cancer.
- To compare the predictive power of quantitative nuclear features against established prognostic factors.
Main Methods:
- Digitization and analysis of fine needle aspiration (FNA) cell samples from 187 invasive breast cancer patients.
- Computer-automated nucleus outlining using "snakes" to extract ten nuclear features (size, shape, texture).
- Statistical analysis and a novel machine learning technique (Recurrence Surface Approximation - RSA) for prognostic evaluation.
Main Results:
- Computer-derived nuclear features demonstrated significant prognostic importance.
- RSA and statistical analyses indicated that quantitative nuclear features are superior to tumor size and lymph node status for predicting outcomes.
Conclusions:
- Objective, computer-based nuclear feature analysis provides a more powerful prognostic tool for invasive breast cancer.
- This approach enhances the accuracy of breast cancer prognostication beyond traditional clinical parameters.