Related Experiment Videos
Automated computerized classification of malignant and benign masses on digitized mammograms
Z Huo1, M L Giger, C J Vyborny
1Kurt Rossmann Laboratories for Radiologic Image Research, Department of Radiology, University of Chicago, IL 60637, USA.
Academic Radiology
|April 2, 1998
Summary
This study developed a computer system to differentiate malignant from benign breast masses. The automated system achieved performance comparable to experienced radiologists, potentially reducing unnecessary biopsies.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Oncology
Background:
- Distinguishing malignant from benign breast masses is crucial for patient management.
- Accurate differentiation can reduce the need for invasive procedures like biopsies.
Purpose of the Study:
- To develop and evaluate an automated computer method for classifying breast masses.
- To estimate the likelihood of malignancy based on extracted lesion features.
Main Methods:
- Digitized 95 mammograms with masses from 65 patients.
- Automatically extracted mass margin and density features.
- Used three automated classifiers and compared their performance to experienced and less experienced mammographers using receiver operating characteristic analysis.
Main Results:
- The computer classification scheme achieved an area under the receiver operating characteristic curve (Az) of 0.94.
- This performance was similar to an experienced mammographer (Az = 0.91) and significantly better than less experienced ones (Az = 0.81).
- At 100% sensitivity, the computer scheme had a positive predictive value of 83%, outperforming experienced (12% higher) and less experienced mammographers (21% higher).
Conclusions:
- Automated computerized classification systems show promise in assisting radiologists.
- These systems can aid in differentiating benign from malignant masses.
- This may lead to a reduction in unnecessary breast biopsies.