Pratama, RM Reza Rizky (2015) Identifikasi sel darah merah bertumpuk menggunakan unsupervised bayesian classification pada citra mikroskopik sel darah. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
5110100181-Undergraduate_Thesis.pdf - Accepted Version Download (3MB) | Preview |
Abstract
Jumlah sel dapat digunakan untuk menentukan jenis penyakit seperti anemia disebabkan kurangnya sel darah merah atau leukimia disebabkan lebihnya sel darah putih, dan lain-lain. Dalam proses ekstrasi sel secara otomatis dari data citra mikroskopik sel darah, salah satu masalah adalah terdapatnya sel bertumpuk yang mengakibatkan ketidakakuratan penghitungan jumlah sel. Pada Tugas Akhir ini diimplementasikan identifikasi sel darah merah bertumpuk dalam citra mikroskopik sel darah dengan unsupervised Bayesian classification.
Pemisahan sel yang bertumpuk akan dimodelkan sebagai suatu cluster. Kemudian digunakan algoritma Expectation-Maximization untuk mendapatkan hasil parameter yang optimal sebagai nilai masukan Bayesian classifier. Cluster validity index digunakan untuk estimasi jumlah sel bertumpuk.
Pada uji coba penghitungan sel darah merah bertumpuk menggunakan metode unsupervised bayesian classification dilakukan perbandingan dengan penghitungan secara manual. Indikator undersegmented dan oversegmented menunjukkan bentuk dari sel darah merah yang tidak tepat bertumpuk. Uji coba metode mampu memberikan hasil cukup memuaskan dengan rata-rata akurasi 87,94%, serta rata-rata
viii
error undersegmented dan oversegmented sebesar 3,82% dan 8,24%.
=================================================================================================
Number of cells can be used to determine disease’s type like anemia caused by a lack of red blood cells or leukemia caused by the excess of white blood cells. When red blood cells are extracted automatically from microscopic images, the presence of stacked cells could lead to inaccuracies in counting the number of cells. This final project identified stacked cells from blood microscopic images for red blood cell count with unsupervised Bayesian classification.
Cluster model is used to separate stacked cells. Expectation-Maximization algorithm was used to approximate the optimal values for input parameters in Bayesian classifier. Cluster validity index is used to estimate the amount of overlapped red blood cells.
Experiments compared the values of red blood cell count from the implemented system with the result of manual counting. The error values of undersegmented and oversegmented indicators show red blood cells which are not precisely overlapped. The experiment results displayed some satisfactory results with the accuracy of 87,94% in addition to undersegmented error 3,82% and oversegmented error 8,24%.
| Item Type: | Thesis (Undergraduate) |
|---|---|
| Additional Information: | RSIf 621.367 8 Pra i |
| Uncontrolled Keywords: | Sel Bertumpuk, Sel Darah Merah, Bayesian Classification, Algoritma Expectation-Maximization(EM) |
| Subjects: | T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques |
| Divisions: | Faculty of Information and Communication Technology > Informatics > 55201-(S1) Undergraduate Thesis |
| Depositing User: | Yeni Anita Gonti |
| Date Deposited: | 20 Feb 2020 02:55 |
| Last Modified: | 20 Feb 2020 02:55 |
| URI: | http://repository.its.ac.id/id/eprint/75066 |
Actions (login required)
 |
View Item |