TotalSegmentator_v2.md

TotalSegmentator V2

Dataset Information

TotalSegmentator is currently the largest publicly available annotated CT segmentation dataset. The first version of the data was released in July 2022, and the dataset underwent a significant update in September 2023. There was a modest increase in both the number of images and the number of annotation categories. The total number of images increased from 1204 to 1228 (only the number of images in the test set was increased), and the number of categories increased from 104 to 117. The current public dataset is divided into 1082 training images, 57 validation images, and 89 test images (the v1 version had 65), all of which are publicly available with annotations. For an introduction to the first version of the data, please refer to the link below. For more detailed information about the data changes, please refer to the following:

Changes/improvements for public dataset:

• Increased number of classes from 104 to 117 (all classes of the total task)
• Removed heart chamber classes
• Improved label quality (see above) (better than before but probably still some errors)
• More meta data per image (scanner, pathology, ...)
• Less intrusive defacing
• No more corrupted files
• Example code for converting to nnU-Net format
• Example code for evaluation
• Same subjects as in v1 for training and validation (we did not publish the additional subjects we used for TotalSegmentator v2 training)
• A few more subjects than in v1 for testing
• NOTE: labels for additional tasks are not included

Dataset Meta Information

Dimensions	Modality	Task Type	Anatomical Structures	Anatomical Area	Number of Categories	Data Volume	File Format
3D	CT	Segmentation	Whole Body	Whole Body	117	1228	.nii.gz

Resolution Details

Resolution Level	spacing (mm)	size
Original Image	(1.5, 1.5, 1.5)	(47, 48, 29)
Medium Resolution	(1.5, 1.5, 1.5)	(241, 231, 231)
High Resolution	(1.5, 1.5, 1.5)	(499, 467, 851)

Total number of image voxels: 317,863 (From the 1228 total cases of the dataset).

Label Information Statistics

Label	Organ	Cases	Coverage	Volume (cm³)	Adjusted Volume (cm³)
1	spleen	886	72.15%	177.28	2542.53
2	kidney_right	782	63.68%	127.44	432.66
3	kidney_left	802	65.31%	126.78	527
4	gallbladder	654	53.26%	18.13	146.8
5	liver	937	76.30%	1363.42	4162.89
6	stomach	905	73.70%	228.41	3174.68
7	pancreas	811	66.04%	59.33	145.81
8	adrenal_gland_right	827	67.35%	3.47	9.88
9	adrenal_gland_left	800	65.15%	4.02	13.27
10	lung_upper_lobe_left	1024	83.39%	781.69	2929.41
11	lung_lower_lobe_left	1006	81.92%	633.56	2136.38
12	lung_upper_lobe_right	894	72.80%	729.36	2304.4
13	lung_middle_lobe_right	881	71.74%	328.26	1806.94
14	lung_lower_lobe_right	991	80.70%	734.69	2412.69
15	esophagus	1032	84.04%	27	176.31
16	trachea	792	64.50%	32.48	79.53
17	thyroid_gland	700	57.00%	12.77	62.27
18	small_bowel	811	66.04%	694.43	2584.29
19	duodenum	768	62.54%	47.06	297.74
20	colon	913	74.35%	535.68	2046.86
21	urinary bladder	571	46.50%	144.21	835.44
22	prostate	343	27.93%	27.06	140.62
23	kidney cyst left	30	2.44%	57.39	527.12
24	kidney cyst right	25	2.04%	26.03	914.43
25	sacrum	616	50.16%	171.23	296.25
26	vertebrae S1	589	47.96%	47.04	99.11
27	vertebrae L5	619	50.41%	62.48	99.82
28	vertebrae L4	636	51.79%	62.66	110.09
29	vertebrae L3	645	52.52%	62.09	140.92
30	vertebrae L2	712	57.98%	55.86	93.21
31	vertebrae L1	801	65.23%	50.86	97.68
32	vertebrae T12	871	70.93%	44.88	108.59
33	vertebrae T11	887	72.23%	41.41	75
34	vertebrae T10	877	71.42%	37.97	85.52
35	vertebrae T9	849	69.14%	33.17	67.92
36	vertebrae T8	796	64.82%	30.04	81.54
37	vertebrae T7	745	60.67%	28.14	60.16
38	vertebrae T6	732	59.61%	25.05	56.8
39	vertebrae T5	733	59.69%	24.31	46.18
40	vertebrae T4	738	60.10%	22.98	37.59
41	vertebrae T3	742	60.42%	21.8	40.2
42	vertebrae T2	724	58.96%	23.04	38.84
43	vertebrae T1	704	57.33%	21.92	41.35
44	vertebrae C7	691	56.27%	13.5	32.54
45	vertebrae C6	556	45.28%	6.1	27.45
46	vertebrae C5	349	28.42%	11.3	25.66
47	vertebrae C4	250	20.36%	12.31	25.07
48	vertebrae C3	229	18.65%	12.75	23.21
49	vertebrae C2	252	20.52%	17.02	28.8
50	vertebrae C1	247	20.11%	12.63	23.22
51	heart	916	74.59%	523.93	1305.97
52	aorta	1078	87.79%	181.52	1566.34
53	pulmonary vein	775	63.11%	21.95	53.63
54	brachiocephalic trunk	725	59.04%	5.18	15.39
55	subclavian artery right	724	58.96%	7.2	23.88
56	subclavian artery left	746	60.75%	8.07	22.28
57	common carotid artery right	718	58.47%	2.94	9.7
58	common carotid artery left	746	60.75%	4.56	13.22
59	brachiocephalic vein left	726	59.12%	11.78	31.58
60	brachiocephalic vein right	728	59.28%	6.02	22.04
61	atrial appendage left	669	54.48%	6.72	26.66
62	superior vena cava	757	61.64%	19.23	42.7
63	inferior vena cava	977	79.56%	57.37	159.91
64	portal vein and splenic vein	830	67.59%	20.74	65.64
65	iliac artery left	651	53.01%	15.62	106.37
66	iliac artery right	644	52.44%	16.05	129.88
67	iliac vena left	636	51.79%	28.89	61.56
68	iliac vena right	629	51.22%	24.17	54.32
69	humerus left	609	49.59%	44.34	256.93
70	humerus right	603	49.10%	49	310.76
71	scapula_left	807	65.72%	85.29	327.81
72	scapula_right	786	64.01%	87.12	160.98
73	clavicle_left	727	59.20%	21.91	62.2
74	clavicle_right	723	58.88%	23.47	53.77
75	femur_left	566	46.09%	176.6	857.8
76	femur_right	556	45.28%	174.98	771.3
77	hip_left	644	52.44%	351.62	584.29
78	hip_right	637	51.87%	351.84	587.02
79	spinal_cord	1183	96.34%	51.55	1564
80	gluteus_maximus_left	595	48.45%	443.69	1017.17
81	gluteus_maximus_right	594	48.37%	461.64	1084.03
82	gluteus_medius_left	625	50.90%	225.88	504.59
83	gluteus_medius_right	613	49.92%	222.88	466.93
84	gluteus_minimus_left	578	47.07%	52.44	99.59
85	gluteus_minimus_right	569	46.34%	60.64	122.36
86	autochthon_left	1133	92.26%	289.49	796.14
87	autochthon_right	1131	92.10%	284.58	808.46
88	iliopsoas_left	836	68.08%	200.18	571.27
89	iliopsoas_right	842	68.57%	188.32	557.65
90	brain	249	20.28%	562.19	1546.77
91	skull	470	38.27%	142.36	1157.69
92	rib_right 4	787	64.09%	15.01	32.8
93	rib_right 3	766	62.38%	12.01	28.6
94	rib_left_1	742	60.42%	9.12	21.1
95	rib_left_2	752	61.24%	10.82	24.73
96	rib_left_3	773	62.95%	11.83	27.64
97	rib_left_4	806	65.64%	14.51	33.91
98	rib_left_5	868	70.68%	15.27	38.9
99	rib_left_6	909	74.02%	17.02	42.46
100	rib_left_7	891	72.56%	18.12	43.8
101	rib_left_8	888	72.31%	15.82	41.45
102	rib_left_9	890	72.48%	15.22	33.71
103	rib_left_10	880	71.66%	12.46	34.54
104	rib_left_11	857	69.79%	8	25.96
105	rib_left_12	790	64.33%	3.16	10.47
106	rib_right_1	738	60.10%	9.47	26.35
107	rib_right_2	751	61.16%	10.86	25.05
108	rib_right_5	838	68.24%	16.27	38.45
109	rib_right_6	869	70.77%	18.09	43.02
110	rib_right_7	851	69.30%	19.32	46.01
111	rib_right_8	859	69.95%	17.06	45.3
112	rib_right_9	860	70.03%	16.03	39.26
113	rib_right_10	866	70.52%	12.63	33.55
114	rib_right_11	866	70.52%	7.78	23.15
115	rib_right_12	785	63.93%	3.07	12.83
116	sternum	959	78.09%	49.9	123.91
117	costal_cartilages	1017	82.82%	114.45	294.52

Visualization

Visualization from official website.

Visualization using ITK-SNAP on large image.

Visualization using ITK-SNAP on small image.

File Structure

The file structure is consistent with TotalSegmentator v1. After decompressing the official zip package, it includes a meta.csv file and several sxxxx subdirectories. In the meta.csv, each row represents an imaging data, listing the picture ID (image_id), age (age), gender (gender), institution code (institute), body part or examination range (study_type), and dataset division (split). These data cover CT scans of multiple body parts from different institutions and are divided into three subsets: training, validation, and testing. Each sxxxx subdirectory contains a segmentation folder and a ct.nii.gz file.

Totalsegmentator_dataset_v2
│
├── meta.csv
│
├── s0000
│   ├── segmentations
│   │   ├── adrenal_gland_left.nii.gz
│   │   ├── adrenal_gland_right.nii.gz
│   │   ├── aorta.nii.gz
│   │   └── ...
│   ├── ct.nii.gz
│
├── s0001
├── s0002
├── ...
└── s1429

Author and Institution

Jakob Wasserthal (University Hospital Basel, Department of Radiology and Nuclear Medicine, Switzerland)

Source Information

Official Website: https://github.com/wasserth/TotalSegmentator

Download Link: https://doi.org/10.5281/zenodo.6802613

Article Address: https://pubs.rsna.org/doi/10.1148/ryai.230024

Publication Date: September 2023

Citation

@article{totalsegmentator,
    author = {Wasserthal, Jakob and Breit, Hanns-Christian and Meyer, Manfred T. and Pradella, Maurice and Hinck, Daniel and Sauter, Alexander W. and Heye, Tobias and Boll, Daniel T. and Cyriac, Joshy and Yang, Shan and Bach, Michael and Segeroth, Martin},
    title = {TotalSegmentator: Robust Segmentation of 104 Anatomic Structures in CT Images},
    journal = {Radiology: Artificial Intelligence},
    volume = {5},
    number = {5},
    pages = {e230024},
    year = {2023}
}

Original introduction article is here.