Skip to content

Latest commit

 

History

History
134 lines (104 loc) · 8.03 KB

BM-BronchoLC.md

File metadata and controls

134 lines (104 loc) · 8.03 KB

BM-BronchoLC

Dataset Information

BM-BronchoLC is a bronchoscopy dataset jointly released by Bach Mai Hospital in Hanoi, Vietnam, and the University of Engineering and Technology, Vietnam National University. The dataset contains endoscopic images from 106 lung cancer patients and 102 non-lung cancer patients, with a total of 2,132 anatomically labeled images and 789 lesion images. The data was collected using an Olympus bronchoscopy system, and video frames were extracted at one frame per second, screened, and annotated by experienced bronchoscopy experts. The annotations include precise localization and classification of 11 anatomical landmarks (e.g., vocal cords, trachea, bronchi) and 7 lesion types (e.g., mucosal erythema, tumors). Each case includes at least 10 high-quality images, all of which have been de-identified to protect patient privacy.

This dataset is significant as it is the first to combine precise localization and recognition of anatomical landmarks and airway lesions in a single bronchoscopy dataset, providing a crucial benchmark for developing AI-assisted diagnostic systems. Evaluations using two mainstream AI models, UNet++ and ESFPNet, demonstrated strong performance on segmentation and classification tasks, achieving over 70% Dice coefficients for anatomical landmark recognition. This opens new possibilities for improving the diagnostic accuracy of bronchoscopy and reducing the limitations of subjective clinical judgment.

Dataset Meta Information

Dimensions Modality Task Type Anatomical Structures Anatomical Area Number of Categories Data Volume File Format
2D Bronchoscopy Recognition Bronchial tube Chest and abdomen 11 + 7 2921 PNG

Resolution Details

Dataset Statistics size
min (480, 480)
median (480, 480)
max (480, 480)

Label Information Statistics

Landmark tags, 11 in total

Number Name Lung Cancer Patients Non-Lung Cancer Patients Lung Cancer Images Non-Lung Cancer Images
1 Vocal cords 60 72 120 149
2 Main carina 86 92 163 191
3 Intermediate 54 68 98 111
4 Right upper lobe 62 86 117 202
5 Right lower lobe 71 85 125 200
6 Right middle lobe 71 84 134 187
7 Left lower lobe 77 88 171 224
8 Left upper lobe 72 77 146 144
9 Right main bronchus 84 91 175 199
10 Left main bronchus 89 93 199 246
11 Trachea 36 49 50 75

Disease tags, 7 in total

Number Name Lung Cancer Patients Non-Lung Cancer Patients Lung Cancer Lesions Non-Lung Cancer Lesions
1 Mucosal congestion 16 19 35 42
2 Soot deposition 6 9 13 40
3 Narrowing 43 6 111 8
4 Bulging mucosal swelling 42 9 128 28
5 Mucosal infiltration 74 5 355 9
6 Blood vessel growth 39 2 111 2
7 Tumor 40 0 183 0

Visualization

Bronchoscope case display.

Left: Endoscopic image / Right: corresponding bronchial segmentation.

File Structure

BM-BronchoLC/
├── Lung_cancer/
│   ├── imgs/
│   │   ├── 00dd8f02-3e56-43e5-a584-e4138063ccf0/
│   │   │   ├── 5286dd59-75b5-461d-a280-da5482c0de18/
│   │   │   │   ├── 004e3a22-bc7f-4ced-88b5-f12488e63405.png
│   │   │   │   ├── 0ae93cda-535b-4ad3-ae9e-191384add4f8.png
│   │   │   │   ├── ...
│   │   │   │   └── (20 more files)
│   │   └── ... (105 similar directories)
│   ├── .DS_Store
│   ├── annotation.json
│   ├── labels.json
│   └── objects.json
└── Non_lung_cancer/
    ├── imgs/
    │   ├── 028ee078-68c3-400e-b7f5-aa2a50805f24/
    │   │   ├── 3dfbe232-049a-482e-9f7a-1a3ffccae90a/
    │   │   │   ├── 0173496d-02e5-461d-b259-f99d85441f3f.png
    │   │   │   ├── 13318412-7bf8-4b8f-8fd4-a9cd98fb6351.png
    │   │   │   ├── ...
    │   │   │   └── (10 more files)
    │   └── ... (101 similar directories)
    ├── .DS_Store
    ├── annotation.json
    ├── labels.json
    └── objects.json

Authors and Institutions

There are many people, the main institutions are Bach Mai Hospital, Hanoi Medical University, and Vietnam National University. You can refer to the article: https://www.nature.com/articles/s41

Source Information

Official Website: https://figshare.com/articles/dataset/BM-BronchoLC/24243670

Download Link: https://figshare.com/articles/dataset/BM-BronchoLC/24243670

Article Address: https://www.nature.com/articles/s41597-024-03145-y

Publication Date: 2023-10

Citation

@article{vu2024bm,
  title={BM-BronchoLC-A rich bronchoscopy dataset for anatomical landmarks and lung cancer lesion recognition},
  author={Vu, Van Giap and Hoang, Anh Duc and Phan, Thu Phuong and Nguyen, Ngoc Du and Nguyen, Thanh Thuy and Nguyen, Duc Nghia and Dao, Ngoc Phu and Doan, Thi Phuong Lan and Nguyen, Thi Thanh Huyen and Trinh, Thi Huong and others},
  journal={Scientific Data},
  volume={11},
  number={1},
  pages={321},
  year={2024},
  publisher={Nature Publishing Group UK London}
}

Original introduction article is here.