The Vision Transformer’s efficacy in handling 3D medical images through self-supervised learning has been elevated with the ‘Mask in Mask’ (MiM) framework. This method not only advances hierarchical representation but also emphasizes anatomical continuity, crucial for medical diagnostics. Key Highlights: - Enhanced multi-level feature representation - Cross-level alignment for anatomical accuracy - Pre-trained on a diverse set of CT images, demonstrating superior segmentation and classification results Opinion: MiM’s methodology sets a new standard for medical image analysis, promoting more nuanced and accurate model training that could greatly benefit clinical applications. The potential to expand to other imaging modalities presents a significant opportunity for impacting patient care.