Automatic Extraction of Soft Tissue Tumor from Ultrasonography Using ART2 Based Intelligent Image Analysis

Background: In this paper, we propose a fully automatic intelligent computer vision based method for extracting soft tissue tumor from ultrasound images. Accurate soft tissue tumor area extraction is vital in developing reliable automatic diagnosing tool from ultrasonography. Methods: The proposed method uses many image processing algorithms including intelligent methods such as fuzzy stretching and ART2 clustering algorithm in forming soft tissue tumor object from ultrasonography. Results: In the experiment, the proposed system successfully extracts the target without human intervention in 55 of 60 cases (91.7% accuracy). 51 of such successful extractions have no larger than 20% discrepancy from human experts' manual extractions in size and 28 cases are highly matched to those of human experts' results (less than 10 % discrepancy in size). Conclusion: Such relatively small discrepancy from human expert's judgement is an evidence of the competence of this work.