Klasifikasi Darah Pada Citra Uji Silang Serasi Menggunakan Pendekatan Deep Learning

Wibawa, Ida Bagus Kade Rainata Putra (2024) Klasifikasi Darah Pada Citra Uji Silang Serasi Menggunakan Pendekatan Deep Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025201235-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2026. Download (5MB) | Request a copy

Abstract

Kecepatan dan akurasi dalam memberikan pelayanan kepada pasien merupakan faktor krusial dalam layanan kesehatan masyarakat. Pelayanan kesehatan, termasuk pelayanan transfusi darah, juga menghadapi tantangan khusus yang memerlukan solusi efektif. Dalam pelayanan transfusi darah, uji silang serasi (pratransfusi) memiliki peran penting untuk mencegah komplikasi pada pasien. Namun, permasalahan utama yang dihadapi adalah proses klasifikasi hasil uji silang serasi yang masih dilakukan secara manual. Hal ini dapat menghambat layanan rumah sakit jika tidak didukung oleh jumlah tenaga kesehatan yang memadai. Oleh karena itu, penelitian ini bertujuan untuk membantu mempercepat pelayanan rumah sakit dan mengalokasikan tenaga kesehatan secara lebih efisien melalui otomatisasi proses klasifikasi hasil uji silang serasi menggunakan Deep Learning. Pendekatan deep learning yang digunakan adalah transfer learning dengan arsitektur pre-trained model yang digunakan meliputi InceptionV3, ResNet50, MobileNetV3Small, EfficientNetB7, dan InceptionResNetV2. Tahapan proses pada penelitian ini adalah preprocessing citra hasil uji silang serasi, augmentasi citra, segmentasi citra hingga masuk ke tahap klasifikasi untuk setiap pre-trained model. Preprocessing citra melibatkan proses cropping citra dan anotasi citra untuk segmentasi. Sementara itu, arsitektur segmentasi yang digunakan adalah U-Net dengan backbone model VGG19. Citra terbagi menjadi trainset dan testset, dengan rasio 4:1. Berdasarkan hasil uji coba, model EfficientNetB7 memberikan performa terbaik dengan hyperparameter Adam sebagai optimizer, learning rate bernilai 0.01, hidden units sebesar 512, dan dropout rate bernilai 0.2. Performa model ini adalah nilai precision sebesar 0.9461, recall sebesar 0.9300, f1-score sebesar 0.9374 dan akurasi sebesar 0.9236 pada testset.
==================================================================================================================================
Speed and accuracy in providing services to patients are crucial factors in public health care. Health services, including blood transfusion services, also face specific challenges that require effective solutions. In blood transfusion services, crossmatch test (pre-transfusion) plays a crucial role in preventing complications in patients. However, the main issue faced is the manual classification process of crossmatch test results. This can hinder hospital services if not supported by an adequate number of healthcare professionals. Therefore, this research aims to help expedite hospital services and allocate healthcare personnel more efficiently through automating the classification process of crossmatch test results using Deep Learning. The deep learning approach used is transfer learning with pre-trained model architectures including InceptionV3, ResNet50, MobileNetV3Small, EfficientNetB7, and InceptionResNetV2. The research process includes preprocessing the crossmatch test result images, image augmentation, image segmentation, and entering the classification stage for each pre-trained model. Image preprocessing involves image cropping and annotation for segmentation. Meanwhile, the segmentation architecture used is U-Net with the VGG19 model as the backbone. The images are divided into a training set and a test set, with a ratio of 4:1. Based on the trial results, the EfficientNetB7 model provides the best performance with hyperparameters including Adam optimizer, a learning rate of 0.01, 512 hidden units, and a dropout rate of 0.2. The performance of this model is a precision value of 0.9461, recall of 0.9300, f1-score of 0.9374, and accuracy of 0.9236 on the test set.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Klasifikasi Citra, Pelayanan Transfusi Darah, Uji Silang Serasi, Transfer Learning, U-Net, Image Classification, Blood Transfusion Service, Crossmatch Testing
Subjects:	T Technology > T Technology (General) > T11 Technical writing. Scientific Writing T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T58.5 Information technology. IT--Auditing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	IDA BAGUS KADE RAINATA PUTRA WIBAWA
Date Deposited:	08 Feb 2024 09:55
Last Modified:	08 Feb 2024 09:55
URI:	http://repository.its.ac.id/id/eprint/106614

Actions (login required)

View Item