Accuracy on Bovine Bone Fracture Detection of Two Dimensional B–Mode Ultrasound Images using Polynomial – Intensity Gradient

The detection of bone fracture uses X-rays or CTscans device typically. These instruments have a negative effect of radiation and need high security for both patients and medical technicians. In this paper, we proposed a framework using the high–order polynomial approach and intensity gradient of two – dimensional B – mode ultrasound images for bone fracture detection. According to the ultrasound probe position, bone scanning process produce curved and flat contour surface. The local phase symmetry and morphology operation is used to extract the bone surface feature from the speckles and other noise. Then, a high order polynomial equation is used to obtain the center mass in the bone area. Two methods, Polynomial Tangent Perpendicular Line (PTPL) and Axis Perpendicular Line method are applied to determine the intensity gradient between adjacent columns based on the center of the mass bone area. These methods are tested to the bovine bone with no – fracture bone, and bone with transverse, oblique and comminuted fractures. Both PTPL and APL methods had 100% accuracy in the detection of fracture occurrence. For estimation width of fracture, the PTPL was more accurate than APL method. In the curved contour bone surface, the PTPL method has 1.14% error with the mean absolute error (MAE) of 0.016 mm. While the APL method has 2.63% error with the MAE of 0.04 mm. Meanwhile, in the flat contour bone surface, the PTPL method has 2.41% error with the MAE of 0.03 mm while the APL method has 3.21% error with the MAE of 0.04 mm.