Summary: Pleural effusion is the accumulation of water in various contents in the gap between the membranes surrounding the outer face of the lung and the inner wall of the chest. This case is commonly encountered in cytopathological evaluation. Therefore, the nuclei detection is an important step in the cytopathological evaluation for the diagnosis of pleural effusion. In recent years, deep learning algorithms have achieved significant success in object detection. In this study, the modern convolutional object detection algorithm, Yolov3, was proposed for the purpose of nuclei detection on cytopathological images of pleural effusion.
For this study by combining the cytopathological images taken from the microscope with panorama method, the database to be used for detection was obtained. Experiments were performed on 80 images containing 11157 nuclei. The proposed method in this study achieved 94.10% accuracy, 98.98% sensitivity and 96.48% F-measure. Compared to other methods used for the similar cases in the literature, this method was found to be 10 times faster. This acceleration provides a significant advantage for real-time computer-aided diagnostic (CAD) applications in digital pathology. Thus, the proposed method can be used as a diagnostic tool by pathologists in digital pathology.
Key Words: Cytopathology, Pleural effusion, Computer-aided diagnosis, Yolov3, Panorama, Nuclei detection, Convolutional Neural Networks. |