Segmentasi Darah pada Citra Uji Silang Serasi dengan Pendekatan Deep Learning

Banjarnahor, Januar Evan Zuriel (2024) Segmentasi Darah pada Citra Uji Silang Serasi dengan Pendekatan Deep Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 5025201210-Undergraduate_Thesis.pdf] Text
5025201210-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 April 2026.

Download (7MB) | Request a copy

Abstract

Kecepatan dan ketepatan dalam melayani pasien adalah faktor penting dalam pelayanan kepada masyarakat. Pelayanan kesehatan, termasuk pelayanan transfusi darah, juga memiliki masalah unik yang memerlukan penyelesaian yang efektif. Evaluasi kepuasan pasien merupakan indikator utama untuk menilai kualitas pelayanan kesehatan. Dalam pelayanan transfusi darah, pemeriksaan tes silang serasi (pratransfusi) sangat penting untuk mencegah komplikasi pada pasien. Namun, masalah utama yang dihadapi adalah pembacaan dan dokumentasi hasil uji silang serasi yang masih dilakukan secara manual. Hal ini dapat menghambat pelayanan rumah sakit jika tidak didukung oleh jumlah tenaga kesehatan yang memadai. Oleh karena itu, penelitian ini bertujuan untuk membantu pelayanan rumah sakit sehingga pekerjaan pencatatan rekam medis sudah tidak manual sehingga dapat mengalokasikan tenaga kesehatan dengan lebih efisien melalui otomatisasi pembacaan dan penyimpanan hasil uji silang serasi dengan model deep learning segmentasi citra yang sudah dibuat. Metode yang dilakukan pada penelitian yang telah dilakukan mencakup pengumpulan data, preprocessing data, pemisahan data, augmentasi data, pembuatan model, pelatihan model, dan evaluasi model. Pengumpulan data dilakukan dengan smartphone. Pemisahan data dibagi menjadi pelatihan, validasi, dan pengujian. Jenis augmentasi yang digunakan adalah kontras, brightness, dan noise. Evaluasi model yang digunakan adalah skor IoU, skor F1, presisi, recall, dan akurasi. Model terbaik berdasarkan uji coba skenario adalah encoder ResNeXt50, decoder Pyramid Attention Network (PAN) dengan hyperparameter batch size 16 dan optimizer Adam dengan learning rate 0.001. Performa model terbaik yang didapatkan adalah skor IoU sebesar 0.8636, skor F1 sebesar 0.9238, presisi sebesar 0.9371, recall sebesar 0.9152, dan akurasi sebesar 0.9811. Hasil pada model mampu merepresentasikan pola reaksi darah pada citra orisinal meskipun terdapat beberapa hasil prediksi model kurang detil sehingga representasi pola reaksi darah pada citra tersebut berubah.
=================================================================================================================================
Speed and accuracy in serving patients are important factors in service to the community. Health services, including blood transfusion services, also have unique problems that require effective solutions. Evaluation of patient satisfaction is a key indicator to assess the quality of health services. In blood transfusion services, cross-match testing (pretransfusion) is essential to prevent complications in patients. However, the main problem faced is the reading and documentation of the cross-matched test results which are still done manually. This can hamper hospital services if not supported by an adequate number of health workers. Therefore, this research aims to help hospital services so that the work of recording medical records is no longer manual so that it can allocate health workers more efficiently through the automation of reading and storing the results of the cross-match test with the deep learning model of image segmentation that has been made. The methods used in the research include data collection, data preprocessing, data separation, data augmentation, model building, model training, and model evaluation. Data collection is done with a smartphone. Data separation is divided into training, validation, and testing. The types of augmentation used are contrast, brightness, and noise. The model evaluation used is IoU score, F1 score, precision, recall, and accuracy. The best model based on the test scenario is ResNeXt50 encoder, Pyramid Attention Network (PAN) decoder with hyperparameter batch size 16 and Adam optimizer with learning rate 0.001. The best model performance obtained is the IoU score of 0.8636, F1 score of 0.9238, precision of 0.9371, recall of 0.9152, and accuracy of 0.9811. The results in the model are able to represent the blood reaction pattern in the original image even though there are some prediction results that are less detailed and representation of the blood reaction pattern in the image changes.

Item Type: Thesis (Other)
Uncontrolled Keywords: Pelayanan transfusi darah, Uji Silang Serasi, Segmentasi Citra, Pyramid Attention Network (PAN), ResNeXt50, Blood Transfusion Service, Crossmatch Testing, Image Segmentation
Subjects: T Technology > T Technology (General) > T57.5 Data Processing
T Technology > T Technology (General) > T58.8 Productivity. Efficiency
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User: JANUAR EVAN ZURIEL BANJARNAHOR
Date Deposited: 09 Feb 2024 07:30
Last Modified: 09 Feb 2024 07:30
URI: http://repository.its.ac.id/id/eprint/106467

Actions (login required)

View Item View Item