chn123/astra-nlst-reports

NLST CT Scan Radiology Reports (Astra CTRG) - 100% COMPLETE

This dataset contains structured radiology reports generated by the Astra Foundation Model (zh-Wang-Med/Astra) for all 2,965 low-dose chest CT series from the National Lung Screening Trial (NLST) cohort, corresponding to the Pillar benchmark evaluation series (min_pixel_data_all_pillar_scores.json).

Final Cohort Breakdown (100% Complete)

• Total Series Evaluated: 2,965 / 2,965 CT volumes (100.0%)
• Cancer Positive (y = 1): 140 / 140 series (100.0%)
• Cancer Negative (y = 0): 2,825 / 2,825 series (100.0%)
• Ground-truth clinical follow-up labels sourced from sybil_validation_series_infos_with_labels.json.
• Aligned with Pillar model ensemble 1-6 year lung cancer risk scores (min_pixel_data_all_pillar_scores.json).

Model & Architecture

• Model: Astra (Astra_CTRG)
• LLM Backbone: Qwen2.5-VL-7B-Instruct
• 3D Visual Encoder: Merlin 3D Clinical ResNet encoder (i3_resnet_clinical_longformer_best_clip)
• Input Pipeline: 100% In-memory Direct DICOM volumetric loader (canonical RAS orientation, zero intermediate disk I/O, foreground cropping, spatial resampling to 224x224x160, HU intensity scaling [-1000, 200]).
• Task: Chest CT Radiology Report Generation (CTRG)
• Report Structure: Detailed anatomical findings for 10 anatomical regions (Abdomen, Bone, Breast, Esophagus, Heart, Lung, Mediastinum, Pleura, Thyroid, Trachea and Bronchie) followed by a clinical Conclusion & Recommendations summary.

Files in Repository

• NLST_2965_Astra_Radyoloji_Raporlari.xlsx: Master Excel file containing all 2,965 generated reports, aligned with patient IDs, ground-truth cancer labels, Pillar risk scores, and execution timings.
• astra_radiology_reports_with_labels.xlsx: Labeled Excel report dataset.
• astra_radiology_reports_with_labels.csv: Full CSV dataset for HF Dataset Viewer and tabular exploration.

Key Column Definitions

• Sira: Sequential series index (1 - 2,965)
• PID: Patient identifier
• Seri_Anahtari: Unique series key (<PID>/<StudyUID>/<SeriesUID>)
• DICOM_Yolu: Filesystem path to the source DICOM series
• Min_Pixel_Degeri: Minimum HU pixel value from CT metadata
• Pillar_Ensemble_Skoru: Predicted 1-to-6 year cancer risk score array from the Pillar ensemble model
• Kanser_Etiketi_y: Ground-truth cancer outcome (1: Cancer diagnosed during follow-up, 0: Negative)
• Censor_Time: Follow-up / censor duration in years
• Kanser_Yil_1 .. Kanser_Yil_6: Cumulative cancer diagnosis status for years 1 through 6
• Mask_Yil_1 .. Mask_Yil_6: Observation mask indicators for years 1 through 6
• Y_Seq: Raw 6-year cumulative cancer indicator vector
• Y_Mask: Raw 6-year observation mask vector
• Radyoloji_Raporu: Complete Astra generated chest CT radiology report text
• Rapor_Karakter_Sayisi: Character count of the generated report
• On_Isleme_Suresi_sn: In-memory DICOM loading and MONAI 3D tensor transformation time (sec)
• Uretim_Suresi_sn: Autoregressive text generation time (sec)
• Toplam_Sure_sn: End-to-end processing time per CT volume (sec)
• GPU_ID: Dedicated hardware worker (cuda:0, cuda:1, or cuda:3, 3x NVIDIA A100 40GB)
• Durum: Processing status (SUCCESS)
• Zaman: Completion timestamp

---

📚 References & Upstream Work

This repository contains experimental evaluations and generated radiology reports derived from upstream foundation models and public clinical cohorts. These works were developed by their respective authors. If you refer to or use the underlying models and datasets in your research, please credit and cite the original publications:

@misc{wang2026astrageneralizablereportgeneration,
  title  = {Astra: a generalizable report generation foundation model for 3D computed tomography}, 
  author = {Zhuhao Wang and Fang Chen and Chaohui Yu and Zihan Li and Yuchao Zheng and Jing Wang and Xuan Yang and Jia Guo and Zhenlu Yang and Xingju Zheng and Yihua Sun and Haojie Han and Xiaoxiao Qin and Zhan Feng and Wenbo Xiao and Chao Zhu and Yuehua Li and Shipeng Zhang and Hao Luo and Yunsong Peng and Fan Wang and Hongen Liao},
  year   = {2026},
  note   = {Preprint}
}

@article{nlst2011reduced,
  title   = {Reduced Lung-Cancer Mortality with Low-Dose Computed Tomographic Screening},
  author  = {{National Lung Screening Trial Research Team}},
  journal = {New England Journal of Medicine},
  volume  = {365},
  number  = {5},
  pages   = {395--409},
  year    = {2011},
  doi     = {10.1056/NEJMoa1102873}
}

@misc{agrawal2025pillar0,
  title  = {Pillar-0: A New Frontier for Radiology Foundation Models}, 
  author = {Kumar Krishna Agrawal and Longchao Liu and Long Lian and Michael Nercessian and Natalia Harguindeguy and Yufu Wu and Peter Mikhael and Gigin Lin and Lecia V. Sequist and Florian Fintelmann and Trevor Darrell and Yutong Bai and Maggie Chung and Adam Yala},
  year   = {2025},
  note   = {Preprint}
}