Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.15822 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908891433926656 |
|---|---|
| author | Liang, Renjie Ma, Yiling Xing, Yang Fan, Zhengkang Pan, Jinqian Sun, Chengkun Li, Li Gong, Kuang Xu, Jie |
| author_facet | Liang, Renjie Ma, Yiling Xing, Yang Fan, Zhengkang Pan, Jinqian Sun, Chengkun Li, Li Gong, Kuang Xu, Jie |
| contents | Automated radiology report generation from 3D CT volumes often suffers from incomplete pathology coverage. We provide empirical evidence that this limitation stems from a representational bottleneck: contrastive 3D CT embeddings encode discriminative pathology signals, yet exhibit severe dimensional concentration, with as few as 2 effective dimensions out of 512. Corroborating this, scaling the language model yields no measurable improvement, suggesting that the bottleneck lies in the visual representation rather than the generator. This bottleneck limits both generation and retrieval; naive static retrieval fails to improve clinical efficacy and can even degrade performance. We propose \textbf{AdaRAG-CT}, an adaptive augmentation framework that compensates for this visual bottleneck by introducing supplementary textual information through controlled retrieval and selectively integrating it during generation. On the CT-RATE benchmark, AdaRAG-CT achieves state-of-the-art clinical efficacy, improving Clinical F1 from 0.420 (CT-Agent) to 0.480 (+6 points); ablation studies confirm that both the retrieval and generation components contribute to the improvement. Code is available at https://github.com/renjie-liang/Adaptive-RAG-for-3DCT-Report-Generation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_15822 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Beyond the Embedding Bottleneck: Adaptive Retrieval-Augmented 3D CT Report Generation Liang, Renjie Ma, Yiling Xing, Yang Fan, Zhengkang Pan, Jinqian Sun, Chengkun Li, Li Gong, Kuang Xu, Jie Computer Vision and Pattern Recognition Automated radiology report generation from 3D CT volumes often suffers from incomplete pathology coverage. We provide empirical evidence that this limitation stems from a representational bottleneck: contrastive 3D CT embeddings encode discriminative pathology signals, yet exhibit severe dimensional concentration, with as few as 2 effective dimensions out of 512. Corroborating this, scaling the language model yields no measurable improvement, suggesting that the bottleneck lies in the visual representation rather than the generator. This bottleneck limits both generation and retrieval; naive static retrieval fails to improve clinical efficacy and can even degrade performance. We propose \textbf{AdaRAG-CT}, an adaptive augmentation framework that compensates for this visual bottleneck by introducing supplementary textual information through controlled retrieval and selectively integrating it during generation. On the CT-RATE benchmark, AdaRAG-CT achieves state-of-the-art clinical efficacy, improving Clinical F1 from 0.420 (CT-Agent) to 0.480 (+6 points); ablation studies confirm that both the retrieval and generation components contribute to the improvement. Code is available at https://github.com/renjie-liang/Adaptive-RAG-for-3DCT-Report-Generation. |
| title | Beyond the Embedding Bottleneck: Adaptive Retrieval-Augmented 3D CT Report Generation |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.15822 |