ShadoNet: A Cell Detection and Classification Framework for Ki-67 Pathology Images.

ShadoNet achieves improved cell detection and classification accuracy in Ki-67-stained histopathology images, enhancing tumor grading for pancreatic neuroendocrine tumors.

• Why it matters: Accurate identification of tumor versus non-tumor cells is crucial for reliable tumor grading, but heterogeneity in Ki-67-stained regions makes this challenging, impacting clinical decision-making.
• What they did: The approach employs a U-Net-style framework called ShadoNet that integrates shape priors with structured regression, using shape masks from the Segment Anything Model and auxiliary loss functions to refine predictions across multiple datasets.
• The result: Incorporating shape information significantly boosts detection and classification performance, enabling more precise tumor cell analysis without requiring detailed boundary annotations, thus supporting better pathology assessments.