Aplikasi Rangkaian Neural HMLP Untuk Saringan Barah Pangkal Rahim Berdasarkan Imej Thinprep
Pap smear test is commonly used as screening test to identify precancerous cells in the cervix. However, it has some limitations due to human and technical errors. To address these limitations, a new technique was proposed known as the ThinPrep. Diagnosis system based on artificial intelligence...
| Main Author: | |
|---|---|
| Format: | Monograph |
| Language: | English |
| Published: |
Universiti Sains Malaysia
2006
|
| Subjects: | |
| Online Access: | http://eprints.usm.my/58552/ http://eprints.usm.my/58552/1/Aplikasi%20Rangkaian%20Neural%20HMLP%20Untuk%20Saringan%20Barah%20Pangkal%20Rahim%20Berdasarkan%20Imej%20Thinprep_Mohd%20%20Izuddin%20Kasim.pdf |
| Summary: | Pap smear test is commonly used as screening test to identify precancerous
cells in the cervix. However, it has some limitations due to human and technical
errors. To address these limitations, a new technique was proposed known as the
ThinPrep. Diagnosis system based on artificial intelligence such as neural network has
been proved in increasing the diagnostic performance. The purpose of this project is
to build cervical cancer diagnosis system using the HMLP network which is trained
using MRPE algorithm. The analysis of neural networks and diagnosis system is built
using Borland C++ Builder software version 6. The diagnosis is done based on
clinical data of image features of ThinPrep test samples. There are 9 image features
were proposed as an input to the HMLP network to classify cervical cell into normal,
LSIL and HSIL cell. The image features were area, blue level, green level, grey level,
red level, intensity, intensity1, perimeter and saturation of cervical cell. Dominant
features analysis bring into play to discover the image features that cause major effect
to the diagnosis. Results show that the dominant image features for this project were
area and perimeter of cervical cell. For overall diagnostic performance, the proposed
diagnosis system based on the HMLP network produced 88.5841% of accuracy. This
proves that the HMLP network has high applicability as intelligent classifiers to
diagnose cervical cancer. |
|---|