ANALISIS TEKSTUR UNTUK DISKRIMINASI MASSA KISTIK DAN NON-KISTIK PADA CITRA ULTRASONOGRAFI
| Main Authors: | , Hari Wibawanto, Drs.MT., , Prof. Adhi Susanto, M.Sc., Ph.D., |
|---|---|
| Format: | Thesis NonPeerReviewed |
| Terbitan: |
[Yogyakarta] : Universitas Gadjah Mada
, 2011
|
| Subjects: | |
| Online Access: |
https://repository.ugm.ac.id/90856/ http://etd.ugm.ac.id/index.php?mod=penelitian_detail&sub=PenelitianDetail&act=view&typ=html&buku_id=53588 |
Daftar Isi:
- An effort to identify cystic as well noncystic masses in ultrasound images, in which 38 samples were of size 21x21 pixels cystic and 89 noncystic, 30 were of size 35x35 pixels cystic and 52 noncystic and 23 were of size 50x50 pixels cystic and 55 noncystic were taken. Each image was tranformed into a grey-level run-length matrix and a greylevel co-occurrence matrix, where 11 and 8 features were extracted, respectively. The merit of these feature in distinguishing the cystic masses from non-cystic ones was tested based on discriminant analysis, using the statistical software package SPSS version 11.5 The results show that an accuracy of 87.3% for 21x21, 91.5% for 35x35, and 94.9% for 50x50 pixels. Further analysis showed that AUROC (Area Under the Receiver Operating Curve) are 0.863, 0.971, and 0,995 for 21x21, 35x35, and 50x50 pixel images, respectively, which confirm that the cystic mass identification scheme has performed sufficiently well. Based on that results, it concluded that texture analysis based-on features extracted from GLC and GLRL matrices can be used to discriminate cystic masses and non-cystic masses with an accuracy from 87,3% to 94,9% depend on size of ROI. Larger ROI resulting in higher accuracy.