MIST

Uses

How it works

Benefits

Comparisons

Highlighted claims

• MIST repositions molecular supervision to the projection layer rather than the encoder, using paired spatial transcriptomics data only during training to construct molecular prototypes. — Bridging the Modality Bottleneck in Pathology MIL through Virtual Molecular Staining
• At inference time, MIST requires no transcriptomics data. — Bridging the Modality Bottleneck in Pathology MIL through Virtual Molecular Staining
• MIST improves 93.75% of model-task configurations over the standard projection layer, with an average gain of 3.5%. — Bridging the Modality Bottleneck in Pathology MIL through Virtual Molecular Staining
• MIST uses sigmoid rather than softmax affinities so that a patch can activate multiple prototypes simultaneously, reflecting mixed cellular programs. — Bridging the Modality Bottleneck in Pathology MIL through Virtual Molecular Staining
• The net parameter increase from MIST is less than 0.09M for seven of eight evaluated architectures. — Bridging the Modality Bottleneck in Pathology MIL through Virtual Molecular Staining
• MIST is reusable across any frozen foundation model and composable with any MIL aggregator because neither the encoder nor the aggregator is modified. — Bridging the Modality Bottleneck in Pathology MIL through Virtual Molecular Staining
• MIST's residual formulation preserves the original morphological patch signal regardless of prototype alignment, analogous to how staining leaves the underlying tissue accessible. — Bridging the Modality Bottleneck in Pathology MIL through Virtual Molecular Staining