Mitigating Biases in Surgical Operating Rooms with Geometry
This addresses bias in AI models for surgical assistance systems, which is crucial for accurately recognizing personalized workflow traits like skill level, though it is incremental as it builds on existing geometric methods for a specific domain.
The paper tackled the problem of deep neural networks learning spurious correlations in surgical operating rooms due to standardized attire, which obscures identifying landmarks and introduces model bias. By using 3D point cloud sequences to encode personnel, they found that geometric representations outperformed RGB models by 12% accuracy in realistic clinical settings, capturing more meaningful biometric features.
Deep neural networks are prone to learning spurious correlations, exploiting dataset-specific artifacts rather than meaningful features for prediction. In surgical operating rooms (OR), these manifest through the standardization of smocks and gowns that obscure robust identifying landmarks, introducing model bias for tasks related to modeling OR personnel. Through gradient-based saliency analysis on two public OR datasets, we reveal that CNN models succumb to such shortcuts, fixating on incidental visual cues such as footwear beneath surgical gowns, distinctive eyewear, or other role-specific identifiers. Avoiding such biases is essential for the next generation of intelligent assistance systems in the OR, which should accurately recognize personalized workflow traits, such as surgical skill level or coordination with other staff members. We address this problem by encoding personnel as 3D point cloud sequences, disentangling identity-relevant shape and motion patterns from appearance-based confounders. Our experiments demonstrate that while RGB and geometric methods achieve comparable performance on datasets with apparent simulation artifacts, RGB models suffer a 12% accuracy drop in realistic clinical settings with decreased visual diversity due to standardizations. This performance gap confirms that geometric representations capture more meaningful biometric features, providing an avenue to developing robust methods of modeling humans in the OR.