A texture analysis approach for objective uniformity evaluation in diagnostic ultrasound imaging: A preliminary study.

Abstract

Ultrasound image uniformity is an important parameter for quality assurance in diagnostic ultrasounds, but it is usually assessed by a qualitative judgement of technicians so its estimation is rough and subjective. In this work a novel method is developed to give an objective measurement of the Bmode image uniformity over the whole field of view or some of its part: the Texture Distribution Analysis Method (TDAM) is based on a segmentation of the Region of Interest, depending on some texture features calculated from co-occurrence matrices. Results on a set of 10 test images with different non-uniformities (Uniformity Image Test Set or UITS) show a good sensitivity and agreement of TDAM with the mean judgment by 5 human observers (TUV): TDAM and TUV uniformity values are coherent for the whole UITS, nevertheless a high uncertainty in uniformity values has been observed (up to 28 percent). Preliminary results look encouraging and more efforts are worth to refine the method.