Akses terbuka Jurnal SINTA 1
A Data-Efficient Image Transformer (DeiT) Approach for Automated Microscopic Blood Cell Classification
Journal of Robotics and Control (JRC) · Vol. 7 No. 4 (2026) · hlm. 3932–3949
- ISSN
- 2715-5072, 2715-5056
- Terbit
- 2026-07-01
- Penerbit
- Universitas Muhammadiyah Yogyakarta
- Bahasa
- eng
Abstrak
Automated microscopic blood cell classification plays a critical role in hematological diagnostics by enabling faster and more consistent cell identification compared to manual examination. However, conventional CNN-based methods often rely on large annotated datasets and fail to efficiently model global contextual dependencies, limiting their scalability and interpretability in clinical settings. To address these challenges, this study proposes a Data-Efficient Image Transformer (DeiT) approach for classifying eight types of normal blood cells from microscopic images. The main contribution of this research is the development of a data-efficient yet interpretable transformer-based framework capable of attaining superior diagnostic precision with minimal annotated data while preserving biological interpretability via visualization-based methods. The proposed model employs a patch-based embedding and self-attention mechanism to extract both localized morphological information and overarching contextual associations between cellular components. Knowledge distillation is integrated to enhance data efficiency, while Grad-CAM and Score-CAM are utilized to visualize the regions most influential in model predictions, ensuring interpretability. Experimental results using 17,092 blood cell images demonstrate excellent performance, achieving 99.52% accuracy, 97.10% recall, 96.42% precision, 99.71% specificity, 96.76% F1-score, and a macro-AUC of 0.998. The DeiT model accurately distinguishes morphologically similar cells such as monocytes and immature granulocytes while focusing on clinically relevant regions in Score-CAM maps. Comparative analysis with recent studies confirms that the proposed approach outperforms several state-of-the-art CNN and hybrid models. In conclusion, the Data-Efficient Image Transformer demonstrates a powerful, interpretable, and resource-efficient solution for automated blood cell classification, offering significant potential for real-world hematologic applications.
Kata kunci
Jaringan kolaborasi para penulis
25 penulis · 68 tautan kepenulisan bersama · 32 lintas organisasi
Arahkan kursor atau ketuk sebuah nama untuk menyorot rekan penulisnya. Klik dua kali untuk membuka halaman penulis.
- Institut Teknologi Dan Bisnis STIKOM Bali
- Rajamangala University of Technology
- Institut Teknologi Sepuluh Nopember
- National Taiwan University of Science and Technology
- Universiti of Malaysia Sabah
- Valaya Alongkorn Rajabhat University
- Afiliasi tidak diketahui
- Organisasi berbeda
- Organisasi sama
- Afiliasi tidak diketahui
- Garis lebih tebal = lebih banyak artikel bersama · lingkaran lebih besar = lebih banyak artikel
Tampilkan sebagai tabel
| Rekan penulis | Organisasi | Artikel bersama |
|---|---|---|
| Dechrit Maneetham | Rajamangala University of Technology | 3 |
| I Komang Agus Ady Aryanto | Rajamangala University of Technology | 2 |
| Wawan Yunanto | National Taiwan University of Science and Technology | 2 |
| Yuri Pamungkas | Institut Teknologi Sepuluh Nopember | 2 |
| Padma Nyoman Crisnapati | Rajamangala University of Technology | 2 |
| Yamin Thwe | Rajamangala University of Technology | 2 |
| I Gusti Ngurah Satria Wijaya | Institut Teknologi Dan Bisnis STIKOM Bali | 1 |
| Riza Wulandari | Institut Teknologi Dan Bisnis STIKOM Bali | 1 |
| Ni Wayan Cahya Ayu Pratami | Institut Teknologi Dan Bisnis STIKOM Bali | 1 |
| I Ketut Putu Suniantara | Institut Teknologi Dan Bisnis STIKOM Bali | 1 |
| Candra Ahmadi | National Taiwan University of Science and Technology | 1 |
| Ni Ketut Dewi Ari Jayanti | Institut Teknologi Dan Bisnis STIKOM Bali | 1 |
| Gde Sastrawangsa | Institut Teknologi Dan Bisnis STIKOM Bali | 1 |
| Ni Wayan Deriani | Institut Teknologi Dan Bisnis STIKOM Bali | 1 |
| Muhammad Rifqi Nur Ramadani | Afiliasi tidak diketahui | 1 |
| Edwin Nugroho Njoto | Afiliasi tidak diketahui | 1 |
| Ratri Dwi Indriani | Afiliasi tidak diketahui | 1 |
| M. N. A. Uda | Universiti of Malaysia Sabah | 1 |
| U. Hashim | Universiti of Malaysia Sabah | 1 |
| Rahmah Yasinta Rangkuti | Institut Teknologi Sepuluh Nopember | 1 |
| Kanittha Nakkliang | Valaya Alongkorn Rajabhat University | 1 |
Afiliasi berasal dari OpenAlex; metadata OAI-PMH jurnal tidak memuatnya. Penulis tanpa ID OpenAlex dicocokkan berdasarkan nama, sehingga dua orang bisa tergabung atau satu orang bisa terpisah.