Year: 2026
Type: conference-paper
Abstract: Three-dimensional Computed Tomography (3D CT) is a cornerstone of precision medicine.Most AI diagnostic models analyze large numbers of CT slices uniformly, treating all slices as equally important.While this has partly accelerated radiologists' workflows, it overlooks that clinically relevant information is often sparsely distributed throughout a volume.Without targeted or weighted processing, fine-grained cues may be missed and substantial computation wasted on diagnostically uninformative slices.We propose a radiologist-simulating framework for selective and efficient 3D CT interpretation.Evaluated on a 3D CT dataset covering eight thoracic lesion types, it was compared with state-of-the-art multimodal large language models such as GPT-4o and supervised visual backbones including ViT and ResNet-50.Using accuracy, F1-score, AUC, and blind radiologist assessment, Screen-CLIP achieved an AUC of 0.87 and F1-score of 0.82, surpassing ViT-Base (AUC: 0.84).For report generation, our method outperformed M3D across all metrics, reaching a BLEU-Avg of 29.03, and achieved the highest average Doctors' Score (6.16/10) in a preliminary human evaluation.
Authors: Kaidi Zhang, Zhiyuan Yan, Gao Cheng, Zhenyang Cai
Institution: Chinese University of Hong Kong, Shenzhen
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W7166821154