Classification of liver lesions by ultrasound images

سال انتشار: 1397
نوع سند: مقاله کنفرانسی
زبان: انگلیسی
مشاهده: 355

نسخه کامل این مقاله ارائه نشده است و در دسترس نمی باشد

استخراج به نرم افزارهای پژوهشی:

لینک ثابت به این مقاله:

شناسه ملی سند علمی:

RSMED01_066

تاریخ نمایه سازی: 30 دی 1397

چکیده مقاله:

Introduction: Texture analysis is considered as a suitable method to increase the diagnosis power of medical images. Present study aims to evaluate the ability of texture analysis for discriminating hemangioma and malignant liver tumors in ultrasound images. Methods and materials: This study was performed on liver ultrasound images (7 hemangioma and 10 malignants). 43 ROIs (Region of Interest) were selected from images. About 270 features were extracted from each ROI. The probability of classification error and average correlation coefficients (POE+ACC) were employed for selection of the best discriminating features. LDA (Linear Discriminating Analysis) and PCA (Principle Component Analysis) were used for discriminating ROIs. Results: LDA and PCA were able to discriminate hemangioma and malignant lesions by misclassification rate of 4.55 and 6.82 respectively. Discussion: both of LDA and PCA had acceptable results. Therefore texture analysis may be a useful method for classification of hemangioma and malignant liver tumor

کلیدواژه ها:

نویسندگان

Milad Zeinali Kermani

MSc, Shafa Hospital, Kerman, Iran

Akbar Gharbali

Department of Medical Physics, Faculty of Medicine, UUMS, Urmia, Iran