Related Experiment Videos
Automatic Gleason grading of prostate cancer using quantitative phase imaging and machine learning.
Tan H Nguyen1, Shamira Sridharan1, Virgilia Macias2
1University of Illinois, Beckman Institute for Advanced Science and Technology, Department of Electrical and Computer Engineering, Quantitative Light Imaging Laboratory, Urbana-Champaign, Illinois, United States.
Journal of Biomedical Optics
|March 31, 2017
Summary
We developed an automatic method using quantitative phase imaging and machine learning for diagnosing prostate cancer biopsies. This approach achieves 82% accuracy in Gleason grading, comparable to human pathologists.
Area of Science:
- Biomedical Engineering
- Computational Pathology
- Medical Imaging
Background:
- Accurate Gleason grading of prostate cancer is crucial for treatment decisions.
- Interobserver variability among pathologists can affect grading accuracy.
- Quantitative, objective methods are needed to improve diagnostic consistency.
Purpose of the Study:
- To develop and validate an automated system for Gleason grading of prostate biopsies.
- To leverage quantitative phase imaging and machine learning for nanoscale tissue analysis.
- To provide an objective metric for prostate cancer diagnosis.
Main Methods:
- Utilized a quantitative phase imaging (QPI) system based on interferometry to capture nanoscale architectural data from unlabeled prostate tissue.
- Employed a random forest classifier to analyze textural features and classify image pixels.
- Integrated morphological and quantitative gland/stroma data with logistic regression for Gleason grade 3 vs. 4 discrimination.
Main Results:
- The automated system achieved an 82% accuracy in discriminating Gleason grade 3 from grade 4 prostate cancer.
- This accuracy is comparable to the interobserver variability observed in human pathologists.
- The method provides quantitative metrics from nanoscale architecture for diagnostic assessment.
Conclusions:
- The proposed approach offers a clinically objective and quantitative method for Gleason grading.
- This automated system has the potential to improve diagnostic accuracy and consistency across laboratories.
- Future work includes corroborating results across instruments and integrating findings into advanced computational algorithms.