Szczegóły publikacji
Opis bibliograficzny
Optimizing brain tumor segmentation through CNN U-Net with CLAHE-HE image enhancement / Shoffan SAIFULLAH, Andiko Putro Suryotomo, Rafał DREŻEWSKI, Radius Tanone, Tundo Tundo // W: ICAI3S 2023 [Dokument elektroniczny] : proceedings of the 2023 1st International Conference on Advanced Informatics and Intelligent Information Systems : Yogyakarta, Indonesia, 29th-30th November 2023 / eds. A. Putro Suryotomo, H. Cahya Rustamaji. — Wersja do Windows. — Dane tekstowe. — [Dordrecht] : Atlantis Press, 2024. — (Advances in Intelligent Systems Research ; ISSN 1951-6851). — e-ISBN: 978-94-6463-366-5. — S. 90-101. — Wymagania systemowe: Adobe Reader. — Tryb dostępu: https://www.atlantis-press.com/article/125997487.pdf [2024-02-03]. — Bibliogr. s. 98-101, Abstr. — Publikacja dostępna online od: 2024-02-02. — S. Saifullah - dod. afiliacja: Universitas Pembangunan Nasional Veteran Yogyakarta. – R. Dreżewski - dod. afiliacja: Universitas Ahmad Dahlan, Yogyakarta
Autorzy (5)
- AGHSaifullah Shoffan
- Suryotomo Andiko Putro
- AGHDreżewski Rafał
- Tanone Radius
- Tundo Tundo
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 151812 |
|---|---|
| Data dodania do BaDAP | 2024-02-19 |
| DOI | 10.2991/978-94-6463-366-5_9 |
| Rok publikacji | 2024 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Creative Commons | |
| Wydawca | Atlantis Press |
| Czasopismo/seria | Advances in Intelligent Systems Research |
Abstract
Accurate segmentation of brain tumors in medical images is paramount for precise diagnosis and treatment planning. In this study, we introduce a robust approach for brain tumor segmentation employing Convolutional Neural Networks (CNNs) with Contrast-Limited Adaptive Histogram Equalization (CLAHE) and Histogram Equalization (HE) preprocessing techniques. We leverage the CNN U-Net architecture, enhanced with CLAHE-HE preprocessing, to achieve high precision in brain tumor segmentation. Our evaluation demonstrates the effectiveness of this approach, revealing substantial improvements in accuracy (reaching up to 0.9982), loss (reducing to 0.0054), Mean Squared Error (MSE, decreasing to 0.0015), Intersection over Union (IoU, increase up to 0.9953), and Dice Score (increase up to 0.9977) during training, validation, and testing phases. Notably, the capacity of our model to generalize effectively is evident through the close alignment of validation performance with training results. These findings underscore the potential of preprocessing techniques in enhancing medical image analysis, with the proposed approach showcasing the promise of revolutionizing brain tumor segmentation, thus contributing to more accurate and reliable diagnoses in clinical settings. Future works may explore innovative preprocessing methods and the application of the proposed approach to other medical image segmentation tasks, which will further advance its capabilities and possible applications areas.
