Akademik Veri Yönetim Sistemi
JOURNAL OF POLYTECHNIC-POLITEKNIK DERGISI, 2024 (ESCI)
Cells, as the fundamental units of life, hold significant importance in clinical and academic studies. Cell counting plays a critical role in disease diagnosis and monitoring. Traditional methods, involving machine-based and manual counting, are costly, errorprone, and time-consuming. Therefore, new methods are being developed to reduce costs and increase accuracy. In this study, cell detection and counting were performed using the BBBC005 dataset. U-Net and V-Net models were separately applied for cell detection, and the U-Net model showed better performance. To improve the results further, the U-Net architecture was enhanced with residual blocks. An ensemble architecture combining the original U-Net and the enhanced U-Net was created, achieving an accuracy rate of 96.15%. In the second phase, cell counting was performed using the scikit-image library. Each cell identified by the model was individually labeled and marked with a distinct color. These visual results enhanced the study's reliability. This innovative approach, combining different models' outputs for better results, offers improvements in cell detection and counting, to the literature.