PseudoCell: Hard Negative Mining as Pseudo Labeling for Deep Learning-Based Centroblast Cell Detection

Abstract

<italic>Background:</italic> Deep learning models for patch classification in whole-slide images (WSI) have shown promise in assisting follicular lymphoma grading. However, these models often require pathologists to identify centroblasts and manually provide refined labels for model optimization. <italic>Objective:</italic> To address this limitation, we propose <italic>PseudoCell</italic>, an object detection framework for automated centroblast detection in WSI, eliminating the need for extensive pathologist&#x0027;s refined labels. <italic>Methods:</italic> <italic>PseudoCell</italic> leverages a combination of pathologist-provided centroblast labels and pseudo-negative labels generated from undersampled false-positive predictions based on cell morphology features. This approach reduces the reliance on time-consuming manual annotations. <italic>Results:</italic> Our framework significantly reduces the workload for pathologists by accurately identifying and narrowing down areas of interest containing centroblasts. Depending on the confidence threshold, <italic>PseudoCell</italic> can eliminate 58.18-99.35&#x0025; of irrelevant tissue areas on WSI, streamlining the diagnostic process. <italic>Conclusion:</italic> This study presents <italic>PseudoCell</italic> as a practical and efficient prescreening method for centroblast detection, eliminating the need for refined labels from pathologists. The discussion section&#x00A0;provides detailed guidance for implementing <italic>PseudoCell</italic> in clinical practice.