SEGMENTASI SEMANTIK CITRA PAP SMEAR MENGGUNAKAN KOMBINASI ARSITEKTUR FULLY CONVOLUTIONAL NETWORK, RESIDUAL BLOCK DAN U-NET (FCRU-NET)

SAGITA, RIRIN and Yahdin, Sugandi and Desiani, Anita (2022) SEGMENTASI SEMANTIK CITRA PAP SMEAR MENGGUNAKAN KOMBINASI ARSITEKTUR FULLY CONVOLUTIONAL NETWORK, RESIDUAL BLOCK DAN U-NET (FCRU-NET). Undergraduate thesis, Sriwijaya University.

	Text RAMA_44201_08011181823106.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (2MB) | Request a copy
	Text RAMA_44201_08011181823106_TURNITIN.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (6MB) | Request a copy
	Text RAMA_44201_08011181823106_0027075803_0011127702__01_front_ref.pdf - Accepted Version Available under License Creative Commons Public Domain Dedication. Download (1MB)
	Text RAMA_44201_08011181823106_0027075803_0011127702_02.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (572kB) | Request a copy
	Text RAMA_44201_08011181823106_0027075803_0011127702_03.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (355kB) | Request a copy
	Text RAMA_44201_08011181823106_0027075803_0011127702_04.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (873kB) | Request a copy
	Text RAMA_44201_08011181823106_0027075803_0011127702_05.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (332kB) | Request a copy
	Text RAMA_44201_08011181823106_0027075803_0011127702_06_ref.pdf - Bibliography Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (440kB) | Request a copy

Abstract

Deep layer of u-net architecture results in increased parameters and network complexity. Residual block has the ability to skip connection a few blocks and take an average score so that it can reduce the complexity of the network with accurate performance, but this architecture is used for image classification. Fully convolutional network architecture (FCN) is a simpler architecture and prevents overfitting due to excessive parameters. In this study there will be a combination of FCN architecture and residual block on u-net architecture to create new, more simple and accurate architecture in the segmentation pap smear image. The results of this architecture obtained 92,20% accuracy performance, sensitivity by 77,74%, 94,23% of spesificity, f1-score by 77,90% and IoU by 66,57% using herlev datassets. The result could be concluded that the architecture advanced has done pap smear image segments and predicts the background well shown with accuracy and specificity above 90%. The architecture can predict the characteristics of the nucleus, the cytoplasm and the thin cytoplasm with quite well viewed from sensitivity values above 70%, and the balance between sensitivity values and specificity is good enough to be seen from the value of F1-score above 70%, but the similarities between the image of the segmentation and the groundtruth are still quite low enough to be seen from IoU score less than 70%. Based on these results, it shows that proposed architecture is capable of pap smear image segmentation.

Item Type:	Thesis (Undergraduate)
Uncontrolled Keywords:	Segmentasi Semantik, Pap Smear
Subjects:	Q Science > QA Mathematics > QA273-280 Probabilities. Mathematical statistics
Divisions:	08-Faculty of Mathematics and Natural Science > 44201-Mathematics (S1)
Depositing User:	Ririn Sagita
Date Deposited:	30 Aug 2023 06:37
Last Modified:	30 Aug 2023 06:37
URI:	http://repository.unsri.ac.id/id/eprint/125953

Actions (login required)

View Item