cover
Article Per Year (5 Year)
All
Home Page
OAI Link
Editorial Team
Contact
Reviewer
Google Scholar
Contact Name
Slamet Riyadi
Contact Email
eist@umy.ac.id
Phone
-
Journal Mail Official
eist@umy.ac.id
Editorial Address
Department of Information Technology Faculty of Engineering, Universitas Muhammadiyah Yogyakarta F3 Building, 2nd Floor Brawijaya Street, Tamantirto, Kasihan, Bantul, Yogyakarta 55183 Indonesia
Location
Kab. bantul,
Daerah istimewa yogyakarta
INDONESIA
Emerging Information Science and Technology
Published by Universitas Muhammadiyah Yogyakarta
ISSN : 27226042     EISSN : 27226050     DOI : https://doi.org/10.18196/eist
Core Subject : Science,
Computer Science & IT
Emerging Information Science and Technology is a double-blind peer-reviewed journal which publishes high quality and state-of-the-art research articles in the area of information science and technology. The articles in this journal cover from theoretical, technical, empirical, and practical research. It is also an interdisciplinary journal that interested in both works from the boundaries of subdisciplines in Information Science and Technology and from the boundaries between Information Science and Technology with other disciplines. EIST is an Open Access Journal to advance sharing science and technology. People have rights to read, download, copy, distribute, print and use with proper acknowledgment and citation. There is no publication fees for authors.
Arjuna Subject : Teknik (semua kategori) - Teknik (Lain-Lain)
Articles 141 Documents
 11   12   13   14   15 
Burn Severity Classification Using Deep Learning and Transfer Learning Saphira, Chairani Fitri; Tahalele, Paul L
Emerging Information Science and Technology Vol. 7 No. 1 (2026): May
Publisher : Universitas Muhammadiyah Yogyakarta

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.18196/eist.v7i1.31493

Abstract

The fast and accurate evaluation of the degree of thermal damage still remains a key factor in the decision-making process but the accuracy of diagnosis is limited by the subjectivity of the clinical knowledge. The current study is aimed at the examination of the deep learning approach towards the automated classification of the skin burns into three degrees using CNNs and transfer learning approaches. A carefully curated dataset of 1,221 burn images was prepared using resizing, normalization, and data augmentation. Additionally, the class weighting approach was used in order to overcome the problems associated with the skewness of the data. The comparison of four different architectures was conducted including CNN, MobileNetV2, EfficientNetB0, and ResNet50 using ImageNet pre-trained weights. Experimental results indicate that ResNet50, tuned using the two-stage training approach, performed best, with test accuracy of 81.81%, precision 0.92, recall 0.90, and F1 score 0.91. The custom CNN model was unable to exhibit stochastic convergence, achieving baseline accuracy of 42.57%, whereas MobileNetV2 and EfficientNetB0 delivered moderate results with 50.89% and 61.65%, respectively. Confusion matrix analysis reveals that the residual learning structure generalized well in all levels of severity, including the important minority class of third-degree burns. The results indicate the dominance of deep residual networks in classifying medical images with insufficient data. The results suggest the use of transfer learning in burn assessment support systems.

Page 15 of 15 | Total Record : 141

 11   12   13   14   15 

Filter by Year

2020 2026


Reset
Filter By Issues
All Issue Vol. 7 No. 1 (2026): May Vol. 6 No. 2 (2025) Vol. 6 No. 1 (2025) Vol. 5 No. 2 (2024): November Vol 5, No 1 (2024): May Vol. 5 No. 1 (2024): May Vol. 4 No. 2 (2023): November Vol 4, No 2 (2023): November Vol 4, No 1 (2023): May Vol. 4 No. 1 (2023): May Vol. 3 No. 2 (2022): November Vol 3, No 2 (2022): November Vol 3, No 1 (2022): May Vol. 3 No. 1 (2022): May Vol. 2 No. 2: November 2021 Vol 2, No 2: November 2021 Vol. 2 No. 1: May 2021 Vol 2, No 1: May 2021 Vol 1, No 4: November 2020 Vol. 1 No. 4: November 2020 Vol. 1 No. 3: August 2020 Vol 1, No 3: August 2020 Vol. 1 No. 2: May 2020 Vol 1, No 2: May 2020 Vol 1, No 1: February 2020 Vol. 1 No. 1: February 2020 More Issue