Where the organs sitENRO
The atlas: organs of many people drawn on one spine.

Loading the atlas…

  1. 1,226 CT scans on one spine

    Where the organs sit: textbook levels against 1,226 CT scans

    Anatomy books place each organ and landmark at a vertebral level or a short range of them. Here each scan is lined up on its own backbone and 11 such statements are checked on the scans that show that part of the spine. The level a book names is the one the scan shows for a median of 45% of people (the median of 11 separate agreement rates), and within one vertebra for 91%.

    Open data: 1,638 hospital CT scans from the TotalSegmentator set (CC BY 4.0); every number is recomputed from them.

  2. How to read the picture

    Bodies lined up on one spine#

    Each scan was lined up on its own vertebrae. One step on the ruler at the left is one vertebra: C1 at the top of the neck, T1 at the first rib, L1 at the waist, S1 at the sacrum. An organ is drawn where the people have it: brighter where more of them share the spot, with lines where 15, 50 and 85% of them have it.

    Amber marks what the books say. Cyan marks what the scans show: the thick bar holds the middle half of people, the thin bar nine in ten, the line the median. The lines on each organ are separate: where 15, 50 and 85% of people have it.

    The 1,226 come from 1,638 adult scans of the TotalSegmentator dataset (its authors describe the first version as scans from one Swiss university hospital) that show enough of the spine to line up. 493 are women and 732 men (1 have no sex recorded); the median age is 65. Scans of people under 18, or with no age recorded (301), are left out of the main figures and analysed apart. CT scans are normally made lying on the back with the breath held, which the dataset does not record. Some of the people have the diseases that bring people to hospital. It is a hospital's population, not a healthy one. The data were rebuilt on the corrected version 3 of the dataset: on the same people no claim moved by more than 3.6 points, so what follows does not depend on the vertebra labels that were fixed (details under "Version 2 against version 3").

  3. What you can feel

    Three landmarks on your own body#

    L4Gray's Anatomy (1918): the top of the hip bones is level with the L4 spinous process
    L4–L533% of 663 people have them above the L4–L5 disc, 26% at it, 41% below

    The hollow at the base of the neck sits at T3 (the book: T2–T3; 69% agree). The lowest edge of the rib cage is at the upper part of L2, where the book says L2–L3 (30%). The top of the hip bones is at the L4–L5 disc.

    The line across the hip bones is the usual way to find L4 from the outside. On these scans it points at L4 for 46% of people 53% have it below L4 and 1 of 663 above it. In a published CT series of 834 patients (Shin 2024) 56% have the crests at or above the L4–L5 disc; here it is 59%.

    The books' L4 fits men and women unequally: it holds for 58% of men and 31% of women. Against the same vertebrae the hip bones of women sit 5 mm lower than those of men, as Cooperstein 2017 found; Shin 2024 found the opposite.

  4. Where the books hold

    The big vessels sit where the books put them#

    The aorta divides within half a vertebra of L4 in 68% of people and the iliac veins join within half a vertebra of L5 in 68%; 99% and 100% are within one vertebra. Vessels that lie against the spine move with it.

    Published series agree for the aorta: L4 in two thirds of 441 MRI scans (Chithriki 2002) and 64% of 146 CT angiograms (Barrey 2013).

  5. Kidneys

    The kidneys sit lower than the books say#

    T12–L3Gray 1918: the upper end level with the top of T12, the lower end with L3
    T12–L1 to L3–L4median top and bottom, 632 people

    The top of the kidney is within StatPearls' T12 for 37% of people, and at the top of T12 as Gray puts it for 22%; the bottom is at L3 for 45%. At both ends about nine in ten are within one vertebra of StatPearls' levels.

    The right kidney lies lower than the left in 63% of 749 pairs; by a median of 1.6 cm where it is, and 0.6 cm over all pairs. Gray 1918 says about 1 cm and "usually"; in the other pairs the left sits lower or level.

  6. Behind the ribs

    Spleen, gallbladder and pancreas against the books#

    The top of the spleen is at T11 (Gray 1918) for 36% of people; the middle 90% runs from upper T9 to T12. The lowest point of the gallbladder is at the lower border of L1 (Gray's reading) for 32% and at L1 itself (Wikipedia's) for 24%; its middle 90% spans upper T12 to L4, the widest range of the 11 levels tested.

    The middle of the pancreas lies between the top of L1 and the bottom of L2, as StatPearls puts it, for 82%: a two-vertebra window, so a looser test than the others.

  7. Heart and lungs

    The heart reaches lower than its levels, and the lungs end near T11#

    Wikipedia, the first place most readers look, puts the heart at T5–T8. A whole heart is about 4.7 vertebrae tall, so it rarely fits inside a four-vertebra range (2 of 598 scans). Its top is inside it in 97% and its midpoint (halfway between top and bottom) in 64%; the medians run from T6 to T11.

    The bottom of the lungs is at T11 at the back. StatPearls puts the lung's posterior border at C7 to T10 (25% end there); Mirjalili 2012 found it next to T12 in 153 CT scans.

  8. A plane through everyone

    What a slice at L1 actually cuts#

    Addison's transpyloric plane runs halfway between the neck hollow and the pubic bone, "at L1"; Wikipedia lists the pylorus, the pancreas, the gallbladder and the hila of the kidneys among the structures on it. Drag the line. The list shows which organs a slice at that level cuts, and in how many people.

    Where the pubic bone is in view (383 scans, the ones that reach the pelvis) the plane falls at the lower part of L1: within L1 for 63%, and between lower L1 and upper L2 for 50% (a smaller share than in the 52 scans of Mirjalili 2012, 75%). Within 3 cm of it lie the bottom of the gallbladder in 44%, the middle of the pancreas in 74%, the middle of the left kidney in 55% and of the right in 47%. All four together: 16.2%. This compares measured proxies (the middle of the kidney stands in for the renal hilum, the bottom of the gallbladder for its fundus), not the named structures themselves.

      Drag the line on the picture, or press an arrow key on its handle.

    • What differs between patients

      What differs between patients: lung size, age, sex#

      Against the same vertebrae, the organs measured here sit up to 4.5 cm lower in the third of people with the larger lungs on the scan than in the third with the smaller (taken within each sex), up to 2.0 cm lower in women than in men, and differ by up to 1.3 cm between people under 50 and over 70. These are comparisons between different people at one moment, not changes within a person. Lung size on the scan partly reflects how much air was in the chest, and partly body size and lung disease; it is known for the 712 scans that show both lungs whole. Taking the median over the rows of the table, the order is lung size, then age, then sex. After the vertebra labels were corrected (version 3 of the data), the gap between the sexes roughly halved: 7.0 mm before, 3.9 mm now on the same people. Pick a group and the picture changes.

      Show the picture for

      Larger vs smaller third of lung sizeMen vs women70 and over vs under 50
      Bottom of the lung4.5 cm lower0.4 cm higher1.2 cm lower
      Where the aorta divides0.6 cm lower0.0 cm higher1.3 cm lower
      Top of the spleen3.9 cm lower0.2 cm higher0.9 cm lower
      Bottom of the gallbladder2.8 cm lower2.0 cm higher0.3 cm higher
      Top of the kidney2.0 cm lower0.3 cm higher0.9 cm lower
      Middle of the pancreas1.8 cm lower1.0 cm higher0.8 cm lower
      Top of the hip bones0.0 cm lower0.5 cm higher0.7 cm higher

      Difference between the group medians, against the same vertebrae.

    • How much positions differ

      Which positions are most predictable from a landmark#

      Measured from the landmark that gave the narrowest spread, half of people lie within the distance shown of the median position: from about a centimetre for the bladder to about two and a half for the stomach.

      Bladder, middle±0.9 cm
      Left adrenal gland±1.4 cm
      Left kidney, middle±1.6 cm
      Heart, bottom±1.6 cm
      Right adrenal gland±1.7 cm
      Pancreas, middle±1.7 cm
      Duodenum, middle±1.8 cm
      Spleen, middle±1.8 cm
      Right kidney, middle±1.9 cm
      Large bowel, middle±2.1 cm
      Small bowel, middle±2.2 cm
      Liver, top±2.2 cm
      Left lung, bottom±2.3 cm
      Liver, bottom±2.3 cm
      Right lung, bottom±2.4 cm
      Stomach, bottom±2.5 cm
      Gallbladder, bottom±2.7 cm

      Half of people lie within this distance of the median position, measured from the landmark (hip bones, rib-cage edge or neck hollow) that gave the narrowest spread.

    • Your own landmarks

      Where yours probably are#

      Choose sex and age. Each row gives where the organ lies in people like you, in centimetres from a landmark you can feel. These are lying-down, breath-held, hospital figures, measured from bone. Palpated hip bones read about a vertebra higher than the imaged ones (Cooperstein 2017), and standing kidneys sit about 3.6 cm lower than lying down (Reiff 1999).

      Sex

      Age

    How this was measured

    Data. The TotalSegmentator dataset (Wasserthal and colleagues, 2023; version 3.0.0 of 10 September 2026, CC BY 4.0): 1,939 clinical CT scans with 117 structures outlined by a neural network and corrected by hand, 1,638 of them of adults. This page uses the outlines and the patients' age and sex, not the images. It was first built on version 2.0.1; the section "Version 2 against version 3" below compares the two.

    Alignment. In each scan the page finds the centre of every whole vertebral body (the front part, without the spinous process) and calls the centre of C1 level 1, T1 level 8, L1 level 20, L5 level 24 and S1 level 25. A height between two centres is a fraction between their levels. A vertebra the scan cuts off is not used. A scan needs at least four whole vertebrae in order: 1,226 of the 1,638 qualify. Version 3 has no label for the body of S1: where the top of the sacrum is in view, the S1 level is placed 19.8 mm below it (a rule fitted on 559 version-2 scans that have both; checked against the real S1 of 488 people who have it in version 2: median error 0.3 mm, spread 3.5 mm, about 0.13 vertebra), which applies to 678 of the 1,226 people.

    What is measured. The top or bottom of an organ is the first or last slice holding at least 0.3% of its voxels, so one stray voxel does not count. The rule matters most for the lung base: on the same lungs the single lowest voxel would give 11% where the rule gives 28%; the other four organ ends move by at most 4 points. A claim is tested only on people whose scan shows that end of the organ whole, which is why the counts differ from row to row.

    Holds. A book's level holds when the measured level is within half a vertebra of it; a range such as T12–L3 runs from the top of its first vertebra to the bottom of its last. Where two books give two readings the table shows both, and the headline takes the better one for the books. That favours the books. Most windows are one vertebra wide; where a book names a range (the notch T2–T3 in Wikipedia, the pancreas L1–L2, the heart T5–T8) the window is wider and the test looser. "Lower part of L1" means 0.1 to 0.4 vertebra below the centre of L1, "upper part" the same above, and a disc is within 0.1 of the gap. The aortic fork is taken as the top of the labelled iliac arteries and the vein junction as the top of the labelled iliac veins.

    The picture. Each organ is drawn from the share of people covered at that height who have it at that spot, in rows of 3 mm and columns of 4 mm. The lines mark where 15, 50 and 85% of people have it; the fill is brighter where more do, with each organ scaled to its own brightest spot so the small ones show. The picture is a projection through the body: organs in front of one another overlap. Vertical distances use the median distance between vertebral centres in these scans, so the proportions are a body's.

    Your landmarks. For each organ the page uses the landmark that gave the narrowest spread, and gives the 10th, 50th and 90th percentile of the distance in centimetres. Nothing is fitted: the numbers are the people's.

    Every textbook level, tested

    For 11 single-level claims the book's vertebra holds for between 25% and 82% of people (median 45%), and within one vertebra for a median of 91%. 4 of the 11 hold for more than half of people.

    Textbook vertebral levels against the scans
    Landmark or organThe booksThe scans: median (middle 90%)What the book statesBook level holdsWithin one vertebraPeople
    Hollow at the base of the neck (jugular notch)Gray 1918: T2–T3 (37%); Dartmouth: T3 (49%); Wikipedia: T2–T3 (69%)T3 (upper T2 – T4–T5)a range of vertebrae69%94%252
    Lowest edge of the rib cageWikipedia: L2–L3 (30%)upper L2 (lower T12 – upper L3)one level30%78%697
    Top of the hip bones (iliac crests)Gray 1918: L4 (46%); StatPearls: L4 (46%)L4–L5 (L4 – L5)one level46%100%663
    Aorta divides into the iliac arteriesGray 1918: L4 (68%); StatPearls: L4 (68%)lower L4 (L3–L4 – L5)one level68%99%666
    Iliac veins join into the vena cavaGray 1918: L5 (68%); StatPearls: L5 (68%)upper L5 (L4 – lower L5)one level68%100%676
    Top of the kidneyGray 1918: T11–T12 (22%); StatPearls: T12 (37%)T12–L1 (lower T11 – upper L2)one level37%89%632
    Bottom of the kidneyGray 1918: L3 (45%); StatPearls: L3 (45%)L3–L4 (lower L2 – upper L5)one level45%91%682
    Top of the spleenGray 1918: T11 (36%); Bergman et al.: T10 (30%)upper T11 (upper T9 – T12)one level36%83%523
    Bottom of the gallbladderGray 1918: L1–L2 (32%); Wikipedia: L1 (24%)L2 (upper T12 – L4)one level32%74%592
    Middle of the pancreasStatPearls: L1–L2 (82%); Gray 1918: L2 (30%)lower L1 (T12 – lower L2)a range of vertebrae82%98%663
    Bottom of the lung (back)StatPearls: T10 (25%)T11 (T9 – lower T12)one level25%71%617
    Heart, wholeWikipedia: T5–T8T6 – T11the organ's whole extent2 of 598 (0.3%) (whole organ inside)–598
    Thyroid gland, wholeStatPearls: C5–T1C6 – T2the organ's whole extent20% (whole organ inside)–249

    How far to trust it

    Held out. Using only the scans with the dataset's institute code J (902 people; version 2 called it I), the page predicted where organs lie in the 324 scans with other institute codes (the dataset does not say what the codes mean). Its 10th to 90th percentile intervals (per sex, with the landmark chosen on the training scans only), which should catch 80%, caught 77% of 12,632 measured positions; the average miss was 2.4 cm and the median miss 1.6 cm. Pooling hides spread: by organ position (top, bottom or middle of each organ) the intervals caught between 52% and 89% (the middle 80% of positions: 63% to 86%), and for the middle patient 79% of their own measured positions fell inside; 14% of patients had all of theirs inside. Positions from one person are correlated, so the pooled figure is not 12,632 independent trials.

    Against published series. The table sets the numbers here beside the ones in the literature. The definitions differ a little (their reading is by eye on reformatted images; this one is a measured centre), so a gap of a few points is expected.

    Published series beside this one
    MeasurePublishedHere
    Top of the hip bonesMirjalili 2012: at L4 in 69% of 108 CT scans. Shin 2024: at or above the L4–L5 disc in 56% of 834 CT scans (26.5% above, 29.8% at it).At or above the L4–L5 disc in 59% of 682 people (33% above, 26% at it); within L4 in 46%. Shin's split is 26.5% above, 29.8% at and 43.7% below; fewer sit at L4 here than in Mirjalili's series.
    Hip bones, by sexCooperstein 2017 (systematic review and meta-analysis): the imaged crests nearest the L4–L5 interspace in women and the L4 spinous process in men, which agrees with this page. Kim 2003 (690 MRI and plain films): median L4–L5. Shin 2024 (834 CT) found them higher in women, which disagrees.Women: the upper part of L5; men: the L4–L5 disc. The book's L4 holds for 58% of men and 31% of women.
    Lowest edge of the rib cageMirjalili 2012: at L2 in 58% of 108 CT scans.Nearest vertebra L2 in 45% of 777 people.
    Aorta dividesChithriki 2002: L4 in 67% of 441 MRI scans. Barrey 2013: L4 in 64% of 146 CT angiograms.L4 in 68% of 699 people.
    Iliac veins joinBarrey 2013: L5 in 44% of 146 CT angiograms.L5 in 68% of 681 people. The books' L5 is the whole of the range from the L4–L5 disc to the L5–S1 disc here, which is wider than a reader's eye.
    Middle of the kidneyMirjalili 2012: the middle of the renal hilum at L2 in 68% (left) and 40% (right) of 108 CT scans.The middle of the whole kidney nearest L2 in 56% (left, 745) and 46% (right, 748).
    Bottom of the lung at the backMirjalili 2012: adjacent to T12 posteriorly in 153 thoracic CT scans.Median T11; nearest vertebra T12 in 30% of 1,832 lungs.
    Transpyloric planeMirjalili 2012: between lower L1 and upper L2 in 75% of 52 CT scans.Between lower L1 and upper L2 in 50% of 383 people; median the lower part of L1.
    Fork of the windpipeMirjalili 2012: below the sternal angle, at T5–T6, in most of 153 CT scans.Median T6; T5–T6 in 80% of 686 people.

    Version 2 against version 3

    What changed. This page was first built on version 2.0.1 of the dataset (1,228 scans). Version 3.0.0 keeps all of them, adds 711 more (278 of them of children or without an age), and its record says that vertebra label mix-ups were fixed, bone labels refined and many small errors in the outlines corrected. The method rests on vertebra labels, so every number was recomputed on version 3 and the two versions were set side by side.

    Same people, new outlines. Of the 999 people aligned in both versions, 91 (9%) have at least one vertebra whose centre moved by more than a centimetre. The median change of every landmark level is zero; for between 3% and 7% of people a given level moved by more than half a vertebra. For none of the 11 claims did the share of people the book fits change by more than 3.6 points (the median of the 11 shares went from 44% to 43%), and the order of the claims barely changed (rank correlation 0.99).

    More people. Version 3 adds 433 adults with a spine in view; with them the page now rests on 1,226 people, and the median is 45%. The added people alone give 48%, with single rows moving further (45% for the lung base against 20% on the same people as before): the gap is not the vertebral ruler: the added adults differ also in distances that use no vertebra at all (the top of the kidney lies 45 mm above the lower edge of the rib cartilage against 30 mm in the other people, and the lungs on the scan hold 2,970 mL against 4,040 mL), so these are different patients and scan protocols. The same 4 claims are above one half in the whole of version 3 as in version 2.

    What did move. The difference between men and women shrank: the median gap between the sexes in the organ heights fell from 7.0 mm to 3.9 mm on the same people (4.1 mm in all adults), and women's organs sit lower in 9 of 11 claims there, against all 11 before. Among the people whose vertebrae were relabelled, the gap between the sexes fell in 7 of 11 landmark levels; that suggests, without proving it, that part of what looked like a sex difference came from mislabelled vertebrae. The gap in age (6.9 mm before, 9.1 mm now) and in lung size (16.3 mm before, 17.6 mm now) did not shrink.

    Children. The 301 scans of people under 18 or without an age were analysed apart (162 with a spine in view): the median share is 48%, and the single rows are too few to compare.

    Share of people for whom the book's level holds, by version of the data (people counted in brackets); where two books differ, each column uses the reading that fits that group better
    Landmark or organVersion 2Same people, version 3Adults added in version 3All adults, version 3 (this page)
    Hollow at the base of the neck (jugular notch)71% (170)73% (172)60% (80)69% (252)
    Lowest edge of the rib cage27% (499)29% (496)32% (204)30% (697)
    Top of the hip bones (iliac crests)46% (485)49% (466)39% (201)46% (663)
    Aorta divides into the iliac arteries67% (488)67% (469)71% (201)68% (666)
    Iliac veins join into the vena cava68% (499)68% (480)66% (200)68% (676)
    Top of the kidney32% (887)32% (877)48% (386)37% (1,255)
    Bottom of the kidney44% (962)43% (955)50% (398)45% (1,345)
    Top of the spleen41% (398)42% (398)43% (127)36% (523)
    Bottom of the gallbladder32% (420)33% (419)28% (177)32% (592)
    Middle of the pancreas86% (472)85% (468)74% (199)82% (663)
    Bottom of the lung (back)20% (968)20% (963)45% (262)25% (1,221)
    Median of the 11 shares (people with a spine in view)44% (1,011)43% (1,006)48% (229)45% (1,226)

    Where the picture misleads

    Blur is disagreement

    A kidney is a smooth bean. The cloud is a thousand kidneys at different heights and widths, drawn on top of each other.

    Lying down, breath held

    Standing, kidneys sit lower: a study of 15 people found 0.5 to 7.5 cm, 3.6 cm on average (Reiff 1999). No level here is a standing level.

    Patients

    The median age is 65 (middle 80%: 42 to 82); children are not in the main analysis, and diseases that enlarge or displace an organ are in. The 349 people with no finding give the same medians to within a quarter of a vertebra.

    The outlines

    The dataset's authors say outline problems, mostly neighbouring vertebrae or ribs mixed up, can occur in roughly 15% of subjects. A fifth lumbar vertebra fused to the sacrum, or a sixth, changes what "L4" means; nearly one adult in five has such a transitional vertebra (Nardo 2012: 18% of 4,636 standing pelvic radiographs). Neither this page nor the books can see that.

    Anatomy, not advice

    This is a description of where organs lie in a group of hospital patients. It is not a guide to any procedure, and no level here should be used to place a needle, make an incision or plan radiation. Individual anatomy varies; imaging, not these figures, decides.

    Sources

    Every book level is quoted from a page linked below, and every published figure comes from the abstract of the paper linked. The per-scan numbers behind the tables are in levels.csv (one row per scan, from the dataset above, with its age and sex).