Research map / Probabilistic and scientific modeling / Molecules and proteins
Auditing Sybil: Explaining Deep Lung Cancer Risk Prediction Through Generative Interventional Attributions
Bartlomiej Sobieski, Jakub Grzywaczewski, Karol Dobiczek, Mateusz Wójcik, Tomasz Bartczak, Szatkowski Patryk, Przemysław Bombiński, Matthew Tivnan, Przemyslaw Biecek
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add pairwise interaction terms to main effects when approximating Sybil over nodules (ablation) | SHNAP (LMPI over nodules) · LUNA25 test split and iLDCT | local fidelity R2 | Improved |
| use main effects only (ablation) | SHNAP · LUNA25 test split and iLDCT | local fidelity R2: R2 approx 1 in most cases | Improved |
| set insertion timestep tau to 0.3 instead of 0 or 1 (ablation) | SDB nodule insertion · expert study, 110 scan pairs | radiologist-judged property preservation | Improved |
| replace nodules with generative in-distribution removal instead of naive perturbation (ablation) | SHNAP · attribution stability | attribution stability | Improved |
| perturb random lung regions disjoint from nodules and attention regions (ablation) | gSHNAP · Sybil, random region control | attribution magnitude | No clear effect |
| insert nodule in upper lobes vs middle/lower lobes | SNAP · 240 patient-nodule combinations | SNAP attribution: p <= 0.009 | Improved |
| insert nodule on left vs right lung | SNAP · 240 patient-nodule combinations | SNAP attribution | No clear effect |
| insert nodule closer to pleural surface | SNAP · Sybil radial sensitivity | SNAP attribution: R2 0.071 -> 0.455 with nodule interactions | Hurt |
| replace attention-based regions (e.g., artifacts, ECG electrodes) with synthesized tissue | gSHNAP · LUNA25 and iLDCT | Sybil risk | Improved |
Closest papers by meaning
- Deep Reinforcement Learning for Early Diagnosis of Lung Cancer (AAAI 2024 · similarity 0.86)
- TumorChain: Interleaved Multimodal Chain-of-Thought Reasoning for Traceable Clinical Tumor Analysis (ICLR 2026 · similarity 0.85)
- LungNoduleAgent: A Collaborative Multi-Agent System for Precision Diagnosis of Lung Nodules (AAAI 2026 · similarity 0.84)
- Rethinking Visual Attribution for Chest X-ray Reasoning in Large Vision Language Models (NeurIPS 2026 · similarity 0.83)
- Cross-Modal Interactive Perception Network with Mamba for Lung Tumor Segmentation in PET-CT Images (CVPR 2025 · similarity 0.83)
- Toward a Vision-Language Foundation Model for Medical Data: Multimodal Dataset and Benchmarks for Vietnamese PET/CT Report Generation (NeurIPS 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
