Xultra - Software for Computer-Assisted Detection of Ovarian Follicles in Ultrasound Images

Coordinators: B. Potočnik and D. Zazula

Algorithm (recognition system) "xultra" is a system designed for automated follicle detection in the ovarian ultrasound images. Algorithm input is an ovarian ultrasound image, as a result is returned a set of regions which are most probable ovarian follicles. Recognition system xultra consist of three major parts: pre-processing (from points 1 do 6), segmentation (from points 7 to 9), and classification (point 10).  Algorithm's pseudo code:

  1. Speckle noise reduction using HRGMF filter.
  2. Edge detection using Kirsch operator.
  3. Binarisation of resulting image with OPTIMAL thresholding method.
  4. Morphological edge thinning.
  5. Heuristic edge filling.
  6. Estimation of ovary position based on histograms of edge pixels along rows and columns.
  7. Segmentation of ovary using OPTIMAL thresholding method.
  8. Labeling (identifying) dark regions.
  9. Describing regions with features (area etc.).
  10. Recognising regions using partial decision based on area, compactness, and a ratio between regions area and its bounding box.

Figure 1 presents a sector of 200x300 pixels extracted from the original ovarian ultrasound image (in 256 grey-levels). The expert readings are denoted with a solid line, while the follicles obtained using algorithm xultra are depicted with dotted line.


Figure 1: Ovarian ultrasound subimage. The obtained follicles using algorithm xultra are denoted with dotted line, the follicle reference positions provided by an expert are depicted with solid line

More about xultra

Xultra with some sample ultrasound images

Papers