Distribution-based deep multiple instance learning for tumor proportion scoring in NSCLC
Quick Answer
The proposed distribution-based deep multiple instance learning (MIL) framework enhances tumor proportion scoring (TPS) in non-small cell lung cancer (NSCLC) by employing a two-model approach: an embedding-extraction network and a MIL model for predicting zero-inflated beta parameters.
Quick Take
The proposed distribution-based deep multiple instance learning (MIL) framework enhances tumor proportion scoring (TPS) in non-small cell lung cancer (NSCLC) by employing a two-model approach: an embedding-extraction network and a MIL model for predicting zero-inflated beta parameters. This method significantly surpasses traditional linear and ridge regression models in accuracy and explainability, addressing challenges in annotating histopathological images.
Key Points
- Utilizes a two-model approach for improved TPS prediction in NSCLC.
- Embedding-extraction network captures histopathological features of patches.
- MIL model predicts zero-inflated beta parameters for overall TPS distribution.
- Outperforms baseline linear and ridge regression methods significantly.
- Addresses the challenge of annotating images with limited expert availability.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Accurate assessment of tumor proportion score (TPS) in non-small cell lung cancer (NSCLC) is critical for treatment planning and prognosis. Key challenges include the tedious manual work required to annotate each slide, combined with the limited number of experts certified for this task. Multiple instance learning (MIL) has proven to be an effective approach for predicting TPS scores at the slide level; however, existing methods struggle with non-expressive (zero class) images. Our approach involves two models: (1) an embedding-extraction and multiclass-classification network that captures the histopathological features of individual patches, and (2) a MIL model that aggregates these embeddings to predict zero-inflated beta (ZIBeta) parameters representing the overall TPS probability distribution for the entire slide. Using only slide-level TPS scores as labels, we demonstrate how this end-to-end framework can leverage a novel distribution-based architecture to improve prediction accuracy and explainability. ZIBeta modeling significantly outperforms baseline linear and ridge regression while capturing expected accuracy through distribution concentration.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Tissues and Organs (q-bio.TO) |
| MSC classes: | 68T07, 92C50 |
| ACM classes: | I.2.6; I.4.7; J.3 |
| Cite as: | arXiv:2606.27579 [cs.CV] |
| (or arXiv:2606.27579v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2606.27579 arXiv-issued DOI via DataCite |
Submission history
From: Witold Dyrka [view email]
[v1]
Thu, 25 Jun 2026 22:11:29 UTC (26,366 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →LLM-Guided ANN Index Optimization for Human-Object Interaction Retrieval
A phase-aware LLM agent optimizes human-object interaction retrieval, outperforming Optuna TPE by 33.3% and VDTuner by 34.2% on the HICO-DET benchmark. This method enhances throughput by 15.3x over UniIR and demonstrates strong transferability across vector database management systems.