Akademik Veri Yönetim Sistemi
9th International Artificial Intelligence and Data Processing Symposium, IDAP 2025, Malatya, Türkiye, 6 - 07 Eylül 2025, (Tam Metin Bildiri)
Acute kidney injury is a clinical condition characterised by a rapid reduction in kidney function, potentially leading to irreversible deterioration. Early and precise diagnosis is crucial for effective patient care and preventing long-term complications. Kidney pathology analysis, traditionally subjective and time-consuming, demands expert proficiency. Deep learning offers a promising solution, automating routine tasks and providing rapid, objective, and reproducible analysis to significantly reduce inter-observer variability and streamline diagnostic workflows. This study evaluates deep learning approaches for acute kidney injury stage classification from pathology images. This methodology successfully leveraged transfer learning, adapting powerful pre-trained models such as VGG16, ResNet50, and EfficientNetB5 for robust feature extraction. Compared to traditional machine learning approaches such as K-Nearest Neighbours, Support Vector Machines, Random Forest, and Artificial Neural Networks, these transfer learning approaches generally exhibit superior performance. Among them, the EfficientNetB5 approach particularly stood out, achieving the best classification metrics across all evaluated categories: F1-score (0.9117), accuracy (0.9118), precision (0.9128), and recall (0.9118). These findings suggest that integrating deep learning into kidney pathology workflows can significantly support pathologists with less manual work, support clinical decision-making, accelerate diagnosis, and facilitate personalised treatment strategies.