A MITOTIC CELL DETECTION APPROACH WITH DEEPLABV3+ AND MOBILENETV2

Nemati N., Hancer E., SAMET R.

Applied and Computational Mathematics, cilt.24, sa.3, ss.349-363, 2025 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 24 Sayı: 3
  • Basım Tarihi: 2025
  • Doi Numarası: 10.30546/1683-6154.24.3.2025.349
  • Dergi Adı: Applied and Computational Mathematics
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Aerospace Database, Communication Abstracts, Metadex, zbMATH, Civil Engineering Abstracts
  • Sayfa Sayıları: ss.349-363
  • Anahtar Kelimeler: DeepLabv3+, Histopathology, Mitosis Detection, Semantic Segmentation, XAI
  • Ankara Üniversitesi Adresli: Evet

Özet

In this study, we integrate the Local Interpretable Model-agnostic Explanations (LIME) with the DeepLabv3+ model using MobileNetv2 as the backbone to enhance explainability in mitosis segmentation. The LIME offers insight into the models decision-making process by identifying the key superpixels that have the most impact on mitosis predictions. Experimental results on the ICPR12 and ICPR14 datasets demonstrate that the LIME effectively identifies biologically relevant regions and helps distinguish true mitotic cells from artifacts. The proposed integration enhances the transparency of deep learning-based segmentation models, making them more interpretable and reliable for clinical applications.