Key Observations

The above discussion has described an approach to the reduction of false-positive region from segmented images containing suspicious ROI. The following key observations can be drawn:

1. Region prefiltering: prefiltering regions based on their area is a quick and simple method to reducing false-positive regions while maintaining similar levels of sensitivity prior to filtering. The area threshold Tarea is defined for a circular region with a diameter of 5 mm. This is a similar value to that used in other studies and is stricter than that used by expert radiologists when interpreting film screen mammograms.

2. Feature extraction: By surveying previous studies, a subset of features for use in the reduction of false-positives regions has been evaluated. These features capture, morphological, gray scale, and texture information about each region. Using an unbiased implementation of PCA, the 316-dimensional feature space is reduced to a 37-dimensional feature space.

3. Sensitivity in the detection of breast lesions: Following evaluation of the FP reduction strategy on 200 abnormal segmented abnormal DDSM mammograms, sensitivity levels dropped by over 8% but the average number of false-positive regions per image drops by approximately 98%. Varying the threshold on the ANN classifier using ROC analysis, the expert radiologist can select a threshold that varies the available sensitivity at the expense of an acceptable number of false-positive regions.

0 0

Post a comment