Buffer-MIL: Robust Multi-instance Learning with a Buffer-Based Approach

Bontempo, Gianpaolo; Lumetti, Luca; Porrello, Angelo; Bolelli, Federico; Calderara, Simone; Ficarra, Elisa

doi:10.1007/978-3-031-43153-1_1

Cited by 2 publications

(1 citation statement)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On the other hand, a few criticisms about the proposed architecture can be highlighted: i) while the criticality-based two-stage MIL approach is well suited for localization-related tasks (such as tumor detection), scenarios like tumor staging/survival prediction may require a deeper analysis across multiple critical regions [53]; ii) our DAS-MIL relies on a separate feature extractor for each target scale. While this is advantageous in terms of overall accuracy, it would require additional training stages for introducing new magnification(s) or adapting those already available for targeting a new task; iii) as currently devised, the patch-level feature extractors are trained in a self-supervised fashion and frozen in the subsequent supervised stages of our pipeline.…”

Section: Discussionmentioning

confidence: 99%

A Graph-Based Multi-Scale Approach With Knowledge Distillation for WSI Classification

Bontempo,

Bolelli,

Porrello

et al. 2024

IEEE Trans. Med. Imaging

Self Cite

View full text Add to dashboard Cite

The usage of Multi Instance Learning (MIL) for classifying Whole Slide Images (WSIs) has recently increased. Due to their gigapixel size, the pixel-level annotation of such data is extremely expensive and timeconsuming, practically unfeasible. For this reason, multiple automatic approaches have been raised in the last years to support clinical practice and diagnosis. Unfortunately, most state-of-the-art proposals apply attention mechanisms without considering the spatial instance correlation and usually work on a single-scale resolution. To leverage the full potential of pyramidal structured WSI, we propose a graph-based multi-scale MIL approach, DAS-MIL. Our model comprises three modules: i) a self-supervised feature extractor, ii) a graph-based architecture that precedes the MIL mechanism and aims at creating a more contextualized representation of the WSI structure by considering the mutual (spatial) instance correlation both inter and intra-scale. Finally, iii) a (self) distillation loss between resolutions is introduced to compensate for their informative gap and significantly improve the final prediction. The effectiveness of the proposed framework is demonstrated on two well-known datasets, where we outperform SOTA on WSI classification, gaining a +2.7% AUC and +3.7% accuracy on the popular Camelyon16 benchmark.

show abstract

Section: Discussionmentioning

confidence: 99%