A Study on Malignant Cell Recognition in Rapid On‐Site Cytological Evaluation

ABSTRACT Rapid on‐site evaluation (ROSE) technique plays an important role in clinical pathology diagnosis, which can significantly improve the early detection rate and diagnostic efficiency of malignant cells. Facing the ROSE scene, we propose a one‐stage target detection model based on ViT and a lightweight backbone (ShuffleNetV2, MobileNetV3). PyQt5 framework and Flask framework are used to construct the software system and realize the online recognition function. The experimental results show that the model performs well under the mAP and FLOPs metrics, has lower time complexity, and higher target detection accuracy than the traditional model, and provides technical support for ROSE.