A worse-case boosting algorithm-based intelligent cervical cancer cell image classification approach
Kunyang Teng, Shaowei Chang, Jiarui Han, Ning Xu, Chen Li · 2025
The significant morphological diversity of cervical cells, coupled with the limitations of training datasets, presents a major challenge for the task of cell classification in intelligent screening. To address the challenge, this paper proposes a new learning algorithm-worse case enhancement algorithm. The key idea is to let the classifier study more information from the worse case data with larger gradient norm than other training data. The method is to dynamically assign more training iterations and greater loss weights to them. This paper selects four classical artificial neural network models and uses the SIPaKMeD cervical cell dataset for experimental verification. The results verify the effectiveness of the algorithm in cervical cell classification tasks.