SUN Colonoscopy Video Database

Update

  • 2022 10/08 : Added supplementary data: Accuracy of the latest approved model (July 2022).
  • 2022 07/11 : Revised notations of polyp location in Table 2.
  • 2020 10/07 : Released SUN Colonoscopy Video Database.

Abstract

SUN (Showa University and Nagoya University) Colonoscopy Video Database is the colonoscopy-video database for the evaluation of an automated colorectal-polyp detection. The database comprises of still images of videos, which are collected at the Showa University Northern Yokohama Hospital. Mori Laboratory, Graduate School of Informatics, Nagoya University developed this database. Every frame in the database was annotated by the expert endoscopists at Showa University.

Summary of database

The SUN database includes 49,136 polyp frames taken from different 100 polyps, which were fully annotated with bounding boxes. Non-polyp scenes of 109,554 frames are also included in this database. The characteristics of the database are summarized in Tables 1-3.

In polyp-exsiting frames, each polyp is annotated with a bounding box as shown in Figure 1. The file formats of images, and bounding boxes are jpeg and a text file, respectively. In the text file, each row represents a bounding box of a polyp, that is, "Fielename min_Xcoordinate,min_Ycoordinate,max_Xcorrdinate,max_Ycoordinate,class_id", where class_id of 0 and 1 represent polyp and non-polyp frames, respectively. Here are an examples:
polyp1_00001.jpg 50,100,150,200,0
polyp1_00002.jpg 120,300,250,600,0
...

Figure 1: Examples of annotated images. The information of the bounding box is provided by a text file besides image files.

Table 1: Characteristics of the SUN database.
Patients registered as SUN database (n = 99)
Sex (Male / Female) 71 / 28
Median Age (IQR) 69 (58 – 74)
Lesions registered as SUN database (n = 100)
Median Size (IQR) mm 5 (3 – 7)
Number of diminutive polyp (≤5mm) 60
Morphology (protruded / flat) 66 / 34
Location (Right / Left / Rectum) 47 / 44 / 8
Pathological diagnosis
Hyperplastic polyp 7
Sessile serrated lesion 4
Low grade adenoma 82
Traditional serrated adenoma 2
High grade adenoma 4
Invasive carcinoma 1

Table 2: Breakdown of polyp samples of SUN database.
ID Number of frames Shape Size Location Pathological diagnosis
1 527 Is 6mm Cecum Low-grade adenoma
2 1,313 Is 18mm Rectum High-grade adenoma
3 292 IIa 3mm Ascending colon Low-grade adenoma
4 80 Is 4mm Sigmoid colon Low-grade adenoma
5 930 IIa 3mm Transverse colon Low-grade adenoma
6 491 IIa 3mm Sigmoid colon Low-grade adenoma
7 315 IIa 6mm Descending colon Low-grade adenoma
8 256 Isp 12mm Sigmoid colon Low-grade adenoma
9 136 Is 4mm Sigmoid colon Low-grade adenoma
10 436 IIa 3mm Transverse colon Low-grade adenoma
11 113 IIa 5mm Descending colon Low-grade adenoma
12 538 Is 5mm Rectum Low-grade adenoma
13 479 Is 5mm Transverse colon Low-grade adenoma
14 1,183 IIa 3mm Sigmoid colon Low-grade adenoma
15 487 Is 5mm Transverse colon Low-grade adenoma
16 199 Is 4mm Transverse colon Low-grade adenoma
17 304 Is 4mm Sigmoid colon Low-grade adenoma
18 243 Is 2mm Sigmoid colon Hyperplastic polyp
19 96 IIa 3mm Transverse colon Low-grade adenoma
20 3159 IIa 3mm Ascending colon Low-grade adenoma
21 100 IIa 3mm Sigmoid colon Low-grade adenoma
22 314 IIa 2mm Ascending colon Low-grade adenoma
23 182 Ip 12mm Ascending colon Low-grade adenoma
24 973 Ip 15mm- Sigmoid colon Low-grade adenoma
25 338 Is 7mm Sigmoid colon Low-grade adenoma
26 370 Is 5mm Descending colon Low-grade adenoma
27 249 Is 5mm Ascending colon Hyperplastic polyp
28 195 Is 2mm Transverse colon Low-grade adenoma
29 377 Isp 13mm Sigmoid colon Low-grade adenoma
30 224 IIa 4mm Sigmoid colon Low-grade adenoma
31 183 Ip 12mm Descending colon Low-grade adenoma
32 981 Ip 15mm- Ascending colon Traditional serrated adenoma
33 594 Is 5mm Sigmoid colon Low-grade adenoma
34 245 Is 3mm Ascending colon Low-grade adenoma
35 1,212 Ip 15mm- Sigmoid colon High-grade adenoma
36 815 IIa 7mm Sigmoid colon Low-grade adenoma
37 448 Is 7mm Transverse colon Low-grade adenoma
38 509 Is 5mm Ascending colon Low-grade adenoma
39 713 IIa 13mm Ascending colon Low-grade adenoma
40 159 IIa 5mm Transverse colon Low-grade adenoma
41 108 IIa 3mm Rectum Low-grade adenoma
42 268 Is 7mm Transverse colon Low-grade adenoma
43 260 Isp 10mm Ascending colon Low-grade adenoma
44 745 IIa 5mm Sigmoid colon Low-grade adenoma
45 383 Is 3mm Ascending colon Low-grade adenoma
46 170 IIa 2mm Transverse colon Hyperplastic polyp
47 705 Is 5mm Transverse colon Low-grade adenoma
48 176 Is 3mm Transverse colon Low-grade adenoma
49 181 IIa 3mm Transverse colon Low-grade adenoma
50 740 Ip 10mm Sigmoid colon Low-grade adenoma
51 1,737 IIa(LST-NG) 15mm- Cecum Low-grade adenoma
52 207 IIa 6mm Sigmoid colon Low-grade adenoma
53 245 Is 4mm Rectum Hyperplastic polyp
54 345 Is 4mm Sigmoid colon Low-grade adenoma
55 700 Is 3mm Ascending colon Low-grade adenoma
56 248 Is 4mm Sigmoid colon Hyperplastic polyp
57 326 Is 5mm Transverse colon Low-grade adenoma
58 267 IIa 6mm Transverse colon Sessile serrated lesion
59 646 Isp 8mm Sigmoid colon Traditional serrated adenoma
60 146 IIa 8mm Transverse colon Low-grade adenoma
61 679 Isp 6mm Ascending colon Low-grade adenoma
62 351 Is 7mm Ascending colon Low-grade adenoma
63 632 Is 7mm Rectum Invasive cancer (T1b)
64 81 IIa 3mm Sigmoid colon Low-grade adenoma
65 222 IIa 3mm Cecum Low-grade adenoma
66 1,685 Is 6mm Sigmoid colon Low-grade adenoma
67 191 IIa 5mm Transverse colon Low-grade adenoma
68 1319 Is 15mm- Rectum High-grade adenoma
69 130 IIa 3mm Descending colon Low-grade adenoma
70 264 Ip 15mm- Sigmoid colon Low-grade adenoma
71 1,021 Is 4mm Ascending colon Low-grade adenoma
72 774 Is 5mm Ascending colon Low-grade adenoma
73 1,285 Is 3mm Cecum Low-grade adenoma
74 276 Isp 5mm Sigmoid colon Low-grade adenoma
75 343 Is 3mm Transverse colon Low-grade adenoma
76 343 Is 3mm Cecum Low-grade adenoma
77 215 Is 4mm Ascending colon Low-grade adenoma
78 267 Isp 12mm Sigmoid colon High-grade adenoma
79 76 Is 4mm Descending colon Low-grade adenoma
80 1,192 Is 10mm Sigmoid colon Low-grade adenoma
81 427 Is 6mm Sigmoid colon Low-grade adenoma
82 111 IIa 3mm Sigmoid colon Sessile serrated lesion
83 795 Isp 13mm Rectum Low-grade adenoma
84 218 Is 5mm Descending colon Low-grade adenoma
85 1,393 IIa 8mm Ascending colon Low-grade adenoma
86 257 IIa 4mm Sigmoid colon Low-grade adenoma
87 454 Is 3mm Cecum Low-grade adenoma
88 249 Is 4mm Ascending colon Low-grade adenoma
89 149 Ip 5mm Descending colon Low-grade adenoma
90 479 Is 10mm Ascending colon Sessile serrated lesion
91 1,061 IIa 13mm Ascending colon Low-grade adenoma
92 391 Is 7mm Descending colon Low-grade adenoma
93 452 Is 7mm Descending colon Low-grade adenoma
94 136 Is 6mm Sigmoid colon Low-grade adenoma
95 606 Isp 8mm Sigmoid colon Low-grade adenoma
96 301 Is 5mm Sigmoid colon Hyperplastic polyp
97 431 IIa 15mm- Cecum Sessile serrated lesion
98 170 IIa 4mm Transverse colon Low-grade adenoma
99 161 Is 5mm Sigmoid colon Low-grade adenoma
100 188 IIa 3mm Rectum Hyperplastic polyp

Table 3: Breakdown of non-polyp samples of SUN database.
ID Number of frames Lenth of each video (seconds)
1 9,961 332.0
2 10,073 335.8
3 7,152 238.4
4 14,635 487.8
5 7,916 263.9
6 17,046 511.4
7 5,636 169.1
8 2,568 85.6
9 9,522 317.4
10 7,086 236.2
11 4,832 161.1
12 6,799 226.6
13 6,328 210.9

[Back to top]

Supplementary data

Accuracy of the latest approved model (July 2022)

After the publication (Misawa M et al. Gastrointest Endosc 2021;93(4):960-967e3), we updated the EndoBRAIN-EYE (CADe). Therefore, we conducted re-analysis using the latest model to show current performance. The following tables show the latest performances for SUN Colonoscopy Video Database.
Table 4: The perfromances of the latest model and previously reported model.
Latest model* Reported model†
Percent
(95% confidence interval)		 n/N Percent n/N
Sensitivity (per-lesion) 98.0 (93.0-99.8) 98/100 98.0 (93.0-99.8) 98/100
Sensitivity (per-frame) 91.5 (91.2-91.7) 44,092/48212 ‡,§ 90.5 (90.2-90.7) 44,472/49140 ‡
Specificity 98.2 (98.1-98.2) 90,068/91,764 ‡,§ 93.7 (93.5-93.8) 88,075/94,039 ‡
*The latest model was regulatory approved in July 2022.
†Refer to the Misawa M et al. Gastrointest Endosc 2021;93(4):960-967e3.
‡The numbers of frames were inconsistent since the database was released with the removal of inappropriate frames after the paper publication.
§The frames which were identified as inappropriate frames by the CADe system were excluded from the analysis.

Table 5: Results of the positive videos (per-frame analysis).
Total number of frames True positive* False negative False positive† Unanalyzable frames‡
48,344 44,902 3,964 156 132
*Number of the frames in which the IoU between the predicted bounding box and ground truth more than or equals to 0.3.
†Number of the frames to which the trained model outputs bounding box but the IoU less than 0.3.
‡Number of the frames which were identified as inappropriate frame by the CADe system.

Table 6: Results of the negative videos (per-frame analysis).
Total number of frames False positive True negative Unanalyzable framese*
109,516 1,696 90,068 17,752
*Number of the frames which were identified as inappropriate frame by the CADe system.

Table 7: Results of the per-polyp analysis of the positive videos.
Number of polyp videos True positive False negative
100 98 2


[Back to top]

Terms of use

Copyright

All intellectual property rights, including copyrights, relating to the information contained in this database are held by Showa University Northern Yokohama Hospital and Mori Lab., Graduate School of Informatics, Nagoya University.

Intended use

This database is available for only non-commercial use in research or educational purpose. As long as you use the database for these purposes, you can edit or process images and annotations in this database. Without permission from Mori Lab., commercial use of this dataset is prohibited even after copying, editing, processing or any operations of this database. Please contact us for commercial use or if you are uncertain about the decision.

Citation

If you clearly indicate the following information, you are allowed to refer to or cite the part of this database. By downloading and using the SUN database you agree to cite this database in any publication based on research in which this database has been used.
  • Development of a computer-aided detection system for colonoscopy and a publicly accessible large colonoscopy video database (with video). Masashi Misawa, Shin-ei Kudo, Yuichi Mori, Kinichi Hotta, Kazuo Ohtsuka, Takahisa Matsuda, Shoichi Saito, Toyoki Kudo, Toshiyuki Baba, Fumio Ishida, Hayato Itoh, Masahiro Oda, Kensaku Mori, Gastrointestinal Endoscopy, Vol. 93, Issue 4, pp. 960-967.e3, 2021. DOI: 10.1016/j.gie.2020.07.060
  • SUN Colonoscopy Video Database. Hayato Itoh, Masashi Misawa, Yuichi Mori, Masahiro Oda, Shin-Ei Kudo, Kensaku Mori, 2020, http://amed8k.sundatabase.org/

Distribution

It is prohibited to sell, transfer, lend, lease, resell, distribute, etc., as it is, or copy, edit, or process this database, in whole or in part.

Personal information

The use of personal information sent to issue the download path of this database is limited to the notifications about this database and usage statistics of this database. Management of information follows the Nagoya University Privacy Policy. Note that this website uses Google Analytics with Cookie for analysis of its accesses. If you want to deny this analysis, please disable Cookie in your webbrowser.

Immunity

Showa University, Nagoya University, and Mori Lab. are not responsible for any damage caused by use and download of this database. We do not guarantee the accuracy, completeness, or usefulness of this database. If you violate any of the above terms, or at the discretion of Showa University, Mori Lab., or Nagoya University, you may not be allowed to use the database even after downloading. Showa University, Mori Lab, Nagoya University may arbitrarily revise this agreement without the approval of the registrant and this data user.

Contact

* If you agree to all the terms of use, please send a request e-mail to hitoh (a t) mori.m.is.nagoya-u.ac.jp.
[Back to top]