[4Pin1] インタラクティブ(2)

[4Pin1-47] Bit-pattern kernel filtering for duplication removing in cell annotation.

〇Toru Nagasaka¹ (1. AI & Eye Co., Ltd.)

Keywords:deep learning, image processing, kernel filtering, cell count, pathological diagnosis

In recent years, deep learning architectures have been shown to achieve good classification performance and a digital histopathology image analysis has been used in clinical settings. A partial image which are sequentially cropped from same cell by grid search may result in double counting. In this work, we constructed bit-pattern kernel filtering algorithm to remove duplication in cell annotation. Using unique combination of bit-pattern kernel (3x3) filter which contract particular ON-pixel pattern, the duplication remove in cell annotation was achieved.