Automatic femur segmentation and condyle line detection in 3D MR scans for alignment of high resolution MR

1. A method for femur segmentation and condyle line detection, comprising:

scanning a knee of a patient with medical imaging equipment to obtain 3D imaging data with such equipment;

processing the obtained 3D imaging data in a digital processor to determine two lines tangent to the bottom of knee condyles in an axial and a coronal plane; and

automatically scanning the patient in a plane defined by the determined lines,

wherein one of the lines is included in an axial slice of the 3D imaging data, the axial slice being determined by calculating, for each axial slice in the 3D imaging data, a first average of probabilities between a left side center of a segmented femur and a left condyle, a second average of probabilities between a right side center of the segmented femur and a right condyle, and a third average probability between the center of the segmented femur and a middle point in an inter-condyle line, obtaining a value of the axial slice by summing the first and second average probabilities and subtracting the third probability from this sum, identifying the axial slices whose values are within a predetermined range, and selecting, from the identified axial slices, the axial slice with the most posterior condyle positions.

2. The method recited in claim 1 wherein the processing comprises:

determining an approximate location of knee;

segmenting the femur of the knee in the determined approximate location;

determining a bottom point on the femur portion on the right side and the left side of the segmented femur in the axial slice and a coronal view to determine the two lines.

3. The method recited in claim 2 wherein the determining an approximate location of the knee comprises determining leg boundaries in two axial planes comprising determining regions which are air and regions which are leg in the two axial planes, one being an upper plane within the femur, and one being a lower plane within the tibia and determining the center of the leg in each of the two selected axial planes using the determined leg boundaries.

4. The method recited in claim 1 wherein the probabilities between a left side center of a segmented femur and a left condyle and the probabilities between a right side center of the segmented femur and a right condyle include random walker probabilities.

5. A method for femur segmentation and condyle line detection, comprising:

scanning a knee of a patient with medical imaging equipment to obtain 3D imaging data with such equipment;

processing the obtained 3D imaging data in a digital processor to determine two lines tangent to the bottom of knee condyles in an axial plane and in a coronal plane; and

scanning the patient in a plane defined by the determined lines,

wherein one of the lines is included in an axial slice of the 3D imaging data, the axial slice being determined by calculating, for each axial slice in the 3D imaging data, a first average of probabilities between a left side center of a segmented femur and a left condyle, a second average of probabilities between a right side center of the segmented femur and a right condyle, and a third average probability between the center of the segmented femur and a middle point in an inter-condyle line, obtaining a value of the axial slice by summing the first and second average probabilities and subtracting the third probability from this sum, identifying the axial slices whose values are within a predetermined range, and selecting, from the identified axial slices, the axial slice with the most posterior condyle positions.

6. The method recited in claim 5 wherein the method uses the obtained 3D imaging data having an initial scan orientation of a leg to generate an image of the leg and determines from the image an approximate location of a knee of the leg and then using the determined location defines a volume of interest segments the femur above the knee.

7. The method recited in claim 6 wherein the method extracts condyle lines of the segmented femur in the axial slice and a coronal image by determining bottom points of the segmented femur.

8. The method recited in claim 7 further including determining a transformation matrix to define a new scan orientation from the extracted condyle lines.

9. The method recited in claim 5 wherein the method uses Hidden Markov Models to approximate the location of a knee.