Comparative Analysis of Vision Transformers and Morphological Approaches for Ki-67 Index Estimation on Histopathologic Images: An Experimental Evaluation
[ X ]
Tarih
2024
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
Ieee
Erişim Hakkı
info:eu-repo/semantics/closedAccess
Özet
This research comprehensively compares two different methodologies for predicting the Ki-67 index- morphological-based analysis and Vision Transformers (ViT). The morphological method focuses on the shape and structural features of tissues and cell structures. On the other hand, Vision Transformers represents an innovative approach developed through the use of attention mechanisms and transformer architectures. ViT offers a different perspective by modeling global context information in recognizing image patterns. This analysis provides a deeper understanding of the accuracy, efficiency, and applicability of current techniques in histopathological image processing, highlighting the potential to advance existing methodologies used in cancer diagnosis. This comparative study aims to evaluate the performance differences between morphological analyses and transformer-based models, identifying the most effective and reliable methods for predicting the Ki-67 index. Experimental analysis, revealed that due to the limited number of labeled data on this domain, traditional morphologic approaches are currently more promising than the vision transformers.
Açıklama
32nd IEEE Signal Processing and Communications Applications Conference (SIU) -- MAY 15-18, 2024 -- Tarsus Univ Campus, Mersin, TURKEY
Anahtar Kelimeler
Ki-67, ViT, Morphology, Segmentation, Cancer
Kaynak
32nd Ieee Signal Processing and Communications Applications Conference, Siu 2024
WoS Q Değeri
N/A
Scopus Q Değeri
N/A
