Professional Documents
Culture Documents
Larici 2002
Larici 2002
important role in the assessment of penetrating Nine patients also underwent chest CT, which Statistical Analysis
thoracoabdominal injuries [15, 16]. Further- was performed before abdominal CT. CT scans of The sensitivity, specificity, positive and negative
more, although several reports [12, 14] have the chest were ordered at the discretion of the predictive values, and accuracy for the helical CT di-
suggested that sagittal and coronal reconstruc- trauma service. Chest CT using a 5-mm collimation agnosis of traumatic diaphragmatic injury were cal-
with a 3-mm reconstruction increment was per- culated for the entire population, for patients with
tions have proven useful in confirming the diag-
Downloaded from www.ajronline.org by Washington State Univ. on 10/03/15 from IP address 134.121.47.100. Copyright ARRS. For personal use only; all rights reserved
formed 30 sec after IV injection of contrast mate- right- versus left-sided injuries, and for those with a
nosis of diaphragmatic injury, little data are rial; the contrast material was administered at a rate penetrating versus blunt mechanism of injury. The
available that specifically evaluate the use of of 3 mL/sec. Only one injection of contrast material sensitivity and specificity of the individual findings
these techniques in trauma cases. The purposes was performed for patients who underwent both of traumatic diaphragmatic injury were also calcu-
of our investigation were to determine the accu- thoracic and abdominal CT examinations. One pa- lated. The paired two-tailed t test was used to com-
racy of helical CT for the diagnosis of diaphrag- tient had dedicated 1-mm helical imaging through pare the average thicknesses of the right and left
matic injury in a patient population with a the lower chest and upper abdomen in addition to diaphragmatic crura of patients with an injured dia-
relatively high proportion of penetrating inju- dedicated thoracic and abdominal CT. Chest CT ex- phragm and those of patients with an uninjured dia-
ries, establish a threshold measurement of the tended from the cervicothoracic junction to the up- phragm for the entire study population as well as for
crus that could be used to distinguish between per abdomen. Because the entire length of the patients with trauma from a blunt mechanism of in-
diaphragm was not always completely covered by jury and those with trauma from a penetrating mech-
an uninjured and an injured diaphragm, and de-
chest CT alone, axial and reformatted data from anism of injury.
scribe an additional sign suggesting diaphrag- both chest and abdominal studies were presented Odds ratios were calculated for the ability of CT
matic injury: active arterial extravasation of for retrospective review. Images were photographed to detect diaphragmatic injuries caused by blunt and
contrast material at the level of the diaphragm. using soft-tissue window settings (level, 40 H; penetrating mechanisms of injury. The odds ratio
width, 440 H). represents the likelihood of CT detecting a diaphrag-
Digital CT data from 20 patients were available, matic injury versus the likelihood of CT suggesting
Materials and Methods thus allowing sagittal and coronal reconstructions to a diaphragmatic injury that was subsequently proven
Study Population be performed. Reconstructed images were calculated incorrect at surgery. These odds ratios were then
The trauma registry at a large urban trauma cen- using 0.7-mm spacing between images with com- compared with the Breslow and Day’s test for ho-
ter was searched for cases of surgically or autopsy- mercially available software (Advantage Windows mogeneity of odds ratios to determine whether CT
proven diaphragmatic injury from March 1998 to 4.0; General Electric Medical Systems). Approxi- performed significantly differently for the diagnosis
March 2001. The search yielded 120 cases. Among mately 40 reconstructed images were generated for of diaphragmatic injury in patients with a blunt ver-
these 120 patients, 23 patients underwent helical CT each of the two planes. sus penetrating mechanism of injury. Statistics were
of the abdomen with or without dedicated chest CT performed using commercially available software
before surgery or before death. An additional two CT Review (SAS 8.2; Statistical Analysis System, Cary, NC).
patients with diaphragmatic injury who underwent The CT studies from these patients were retro-
preoperative CT were identified at a collaborating spectively reviewed by five thoracic radiologists
institution, yielding a total study population of 25 who were unaware of the surgical or autopsy find-
Results
patients with proven diaphragmatic injury. ings. The reviewers were aware only of the context Associated abdominal injuries were
The average age of the study population was 36.2 of the study. Reviewers specifically evaluated the present in 14 (56%) of the 25 patients with
years (range, 17–82 years). Seventeen patients sus- imaging studies for an abnormally elevated hemidia- diaphragmatic injury.
tained right-sided injuries, and eight patients had phragm, diaphragmatic discontinuity, visceral herni- The sensitivity, specificity, positive and neg-
left-sided injuries; there were no bilateral injuries. ation with or without the collar sign, an abnormally ative predictive values, and accuracy for the
The mechanism of diaphragmatic injury was pene- thick-appearing crus, the “dependent viscera” sign, detection of diaphragmatic injuries on helical
trating trauma (i.e., gunshot wound or stab wound) and any other finding that could suggest diaphragm
CT were 84% (21/25), 77% (17/22), 81% (21/
in 14 patients and blunt trauma (i.e., injury in a mo- injury. Sagittal and coronal reformations were re-
tor vehicle crash, struck by a car, or fall from a viewed contemporaneously with axial images. Re-
26), 81% (17/21), and 81% (38/47), respec-
height) in 11 patients. viewers described each diaphragm as injured, tively. The sensitivity and specificity of helical
An additional 22 patients who sustained either uninjured, or unknown and the reconstructed images CT for penetrating injury were 86% (12/14)
blunt (n = 14) or penetrating (n = 8) trauma, had sur- for each case as either helpful or not helpful in cor- and 79% (11/14), respectively, and those for
gically confirmed intact diaphragms, and had under- roborating the diagnosis based on the axial images. blunt injury were 82% (9/11) and 75% (6/8).
gone preoperative CT were randomly chosen and The reviewers understood that a diagnosis of “un- The odds ratio for the CT detection of blunt di-
included as the control group, bringing the total known” would prompt surgical exploration to defin- aphragmatic injury was 13.5 (CI, 1.47–123)
study population to 47 patients. itively exclude diaphragmatic injury. The presence and that for penetrating injury was 22.0 (CI,
or absence of associated injuries was also noted. 3.1–157). Such large odds ratios imply a
CT Scans strong correlation between the CT diagnosis of
All patients underwent helical CT (CTi; General Diaphragm Measurements traumatic diaphragmatic injury and surgical
Electric Medical Systems, Milwaukee, WI) of the ab- For the entire study population, the thickness of confirmation of diaphragmatic injury. The
domen and pelvis after receiving water-soluble oral the diaphragmatic crura was measured in an at- ability of CT to detect diaphragmatic injury on
contrast material (Gastrografin [meglumine diatri- tempt to establish a value that would enable radiol-
the basis of the mechanism—blunt or penetrat-
zoate]; Bracco Diagnostics, Princeton, NJ) and IV ogists to distinguish between normal and injured
ing—did not differ significantly ( p = 0.747).
contrast material (Omnipaque 300 [iohexol]; Ny- diaphragms. The diaphragm crura were measured
comed, Princeton, NJ). Scanning with a 7-mm colli- by drawing perpendicular lines bisecting one an- The sensitivity and specificity in detecting
mation and a 7-mm reconstruction interval was other through the center of the spinal canal at the diaphragmatic injuries in patients with a right-
performed 70 sec after IV injection of contrast mate- level of L1 and then measuring the thickness of the versus left-sided injury and in those with a
rial at a rate of 2 mL/sec. Imaging began at the lung crus along a line drawn at a 45° angle to the inter- penetrating versus blunt mechanism of injury
bases, above the diaphragm, in all patients. section of these two lines. as well as the individual helical CT findings of
diaphragmatic injury are presented in Table 1. Sensitivity and Specificity of Helical CT for Diagnosis of Diaphragmatic
The most sensitive finding (52%) for the de- TABLE 1
Injury Based on Side of Injury, Mechanism of Injury, and Imaging Findings
tection of diaphragmatic injury was the re-
Sensitivity Specificity
cently described dependent viscera sign [2] Variable
(Fig. 1). Direct visualization of diaphragmatic % No. % No.
Downloaded from www.ajronline.org by Washington State Univ. on 10/03/15 from IP address 134.121.47.100. Copyright ARRS. For personal use only; all rights reserved
injury (Fig. 2) and the “thick crus” sign [18] Side of injury
(Fig. 3) were both 36% sensitive. The depic-
Right 79 13/17 82 18/22
tion of a missile or puncturing instrument tra-
Left 100 8/8 95 21/22
jectory was the most sensitive finding in
patients with a penetrating diaphragmatic in- Mechanism of injury
jury (Fig. 4). Active arterial extravasation of Blunt 82 9/11 75 6/8
contrast material near the diaphragm (Fig. 2) Penetrating 86 12/14 76 11/14
suggested diaphragmatic injury in two patients Helical CT finding
and was the only sign of diaphragmatic injury Collar sign 24 6/25 100 22/22
in one of these two patients. For the entire pop- Direct visualization of injury 36 9/25 95 21/22
ulation, the collar sign was the most specific Thick crus 36 9/25 77 17/22
(100%) manifestation of diaphragmatic injury Herniation without collar sign 8 2/25 95 21/22
(Fig. 5), although the trajectory of the injury in Trajectory 36 5/14 100 22/22
patients with penetrating injury was also 100%
Active extravasation of contrast material 8 2/25 100 22/22
specific (Fig. 4).
Dependent viscera sign 52 13/25 71 9/22
False-negative interpretations (4/25) oc-
curred most commonly in patients with other
injuries abutting the injured diaphragm, such
as a pleural effusion (Fig. 6) or perisplenic he- sifying the cases according to mechanism of ologists at major trauma centers will frequently
matoma (Fig. 7). injury, we also found that the mean thickness encounter patients with diaphragmatic injuries
Reviewers indicated that diagnostic confi- of the diaphragmatic crura was not signifi- [19]. Diaphragmatic injuries are commonly as-
dence of a true-positive finding was increased cantly greater in patients with an injured dia- sociated with other severe thoracoabdominal in-
by the sagittal and coronal reconstructions in phragm than in those with an uninjured juries [4, 8, 20], and the overall morbidity and
three of the 20 true-positive cases (Fig. 5). Dia- diaphragm (Table 2). mortality associated with traumatic diaphrag-
phragmatic injury was incorrectly suggested matic injuries may both exceed 30% [2, 21–24];
on reconstructed images in two cases (Fig. 8). this mortality rate reflects the frequency and se-
For the entire study population, the mean Discussion verity of other organ injuries associated with di-
thickness of the diaphragmatic crura (right and Diaphragmatic injuries occur in only 5–10% aphragmatic tears rather than the diaphragmatic
left) was not significantly greater in patients of patients with severe thoracoabdominal injury injury itself. The diagnosis of diaphragmatic in-
with an injured diaphragm than in those with [1, 3, 4], but the large number of trauma cases jury usually necessitates prompt surgical repair
an uninjured diaphragm (Table 2). When clas- each year in the United States implies that radi- [5, 6, 22]. If the diagnosis is missed or delayed,
Fig. 1.—30-year-old man involved in motor vehicle collision. Axial CT image Fig. 2.—17-year-old boy with right upper quadrant stab wound. Axial CT image reveals
shows bowel (arrows) is resting against posterior ribs in left lower hemithorax. discontinuity of right hemidiaphragm (curved arrow), which is diagnostic of diaphrag-
This finding represents “dependent viscera” sign. Rupture of left hemidiaphragm matic injury. Active arterial extravasation of contrast material (straight arrow) indi-
was surgically confirmed. cates that injury is in close proximity to diaphragm.
Fig. 3.—60-year-old man involved in motor vehicle collision. Axial CT image shows Fig. 4.—17-year-old girl with right upper quadrant gunshot injury. Axial CT image re-
subjectively thickened right diaphragmatic crus (arrows), suggesting injury. Dia- veals subcutaneous emphysema in superficial tissues of right flank (straight arrow).
phragmatic rupture was confirmed at surgery. Linear, irregular low-attenuation area in liver (curved arrow) is consistent with lac-
eration from missile. Extrapolating missile trajectory indicates that projectile must
have traversed diaphragm. Diaphragmatic injury was confirmed at surgery. Streak
artifact emanates from bullet fragment in posterior soft tissues of thorax.
the pressure difference between the thorax and conventional CT, particularly the ability to per- of diaphragmatic injury, with sensitivities and
the abdomen tends to favor enlargement, rather form rapid volumetric acquisitions through the specificities ranging from 0% to 61% [2, 3, 9–
than healing, of the injury, which results in pro- area of interest using a narrow collimation. 13, 25, 26] and from 76% to 99% [2, 3, 10, 12,
gressive herniation of the abdominal contents Such techniques also allow high-quality sagit- 13, 26], respectively. Several CT findings sug-
into the thorax [3, 6]. tal and coronal reformations to be performed. gestive of diaphragmatic injury have been de-
CT has the major advantages of being These reformations have been shown to be scribed, including apparent elevation of a
widely available, rapid, noninvasive, and accu- beneficial in experimental models and in prac- hemidiaphragm, herniation of the abdominal
rate for the assessment of the numerous injuries tice [3, 12–14]. contents into the thorax (with or without waist-
that are often associated with diaphragmatic in- Several studies have evaluated the utility of like constriction of the herniated viscus, the so-
juries. Helical CT offers major advantages over conventional and helical CT for the evaluation called collar sign) [3], direct visualization of
A B
Fig. 6.—24-year-old man with right upper quadrant stab wound. Axial CT image re- Fig. 7.—65-year-old man involved in motor vehicle collision. Axial CT image ob-
veals moderate-sized right pleural effusion (asterisk) but does not reveal direct ev- tained with 1-mm collimation reveals evidence of splenic parenchymal injury
idence of diaphragmatic injury. Large effusion may obscure direct visualization of (curved arrow). Curvilinear high-attenuation structure (straight arrows) posterior
diaphragmatic injury that was proven at surgery. to spleen was thought to represent injured diaphragm, but no evidence of diaphrag-
matic injury was found at surgery.
diaphragmatic discontinuity, an abnormally One of the purposes of our study was to ex- phragmatic rupture from blunt trauma include
thickened diaphragmatic crus [18], evidence of amine the utility of helical CT for the detection rapid, large increases in intraabdominal pressure
penetrating injury with a trajectory projected of diaphragmatic injury in a patient population or shear stress [3, 6], whereas the injury mecha-
to involve the diaphragm [15, 16, 27], and the with a predominance of penetrating injuries nism from a penetrating injury is related to the
recently described dependent viscera sign [2]. (gunshot or stab wounds) as opposed to patients missile or stabbing weapon itself. These differ-
The latter manifests as the cranial aspects of with blunt trauma injuries. Most CT studies of ences in mechanism raise the possibility that the
the abdominal viscera (the liver on the right diaphragmatic injuries to date have focused on average extent of injury to the diaphragm may
and stomach or bowel on the left) resting patients who sustained blunt trauma [3, 12], but differ in patients who sustain blunt versus pene-
against posterior ribs because of the loss of di- these results may not be generalizable to patients trating trauma [16, 28]. Many diaphragmatic in-
aphragmatic support after injury [2]. with penetrating injuries. Mechanisms of dia- juries related to blunt trauma are large, often
A B
Mean Thickness of Similar to Killeen et al. [12], we found that presence of an injury on the basis of a single
Diaphragmatic Crura: helical CT performed better for left-sided inju- CT finding.
TABLE 2
Injured Diaphragm Versus ries (100% sensitivity) than for right-sided inju- We found sagittal and coronal reformations
Uninjured Diaphragm ries (76% sensitivity). Other investigators have added little to the reviewers’ ability to diagnose
found that left-sided injuries are more readily diaphragmatic injury. In contrast, Killeen et al.
Downloaded from www.ajronline.org by Washington State Univ. on 10/03/15 from IP address 134.121.47.100. Copyright ARRS. For personal use only; all rights reserved
Mean Thickness of
Group Crus (mm) p detected on helical CT [3, 12], perhaps in part [12] found that reformatted images aided in the
because of the relatively greater tissue contrast diagnosis of diaphragm injury, and experimen-
Injured Uninjured
between fat in the left upper quadrant and the tal investigations reported that sagittal reforma-
Entire study adjacent left hemidiaphragm than the poor tis- tions were 92% sensitive for the diagnosis of
Right 7.53 5.83 0.326 sue contrast between the right hemidiaphragm diaphragm injury and were diagnostically supe-
Left 6.27 4.33 0.278 and the adjacent liver. These differences in tis- rior to both coronal and axial reformations [13,
Blunt trauma sue contrast may enhance direct visualization of 14]. In our series, reformatted images increased
Right 10 6 0.314 diaphragmatic injuries. diagnostic confidence for only five (20%) of the
Left 5.75 4.25 0.182 The most sensitive sign of diaphragmatic in- 25 true-positive cases; the sagittal images were
jury in our study was the newly described depen- considered the most useful plane in three of
Penetrating trauma
dent viscera sign. This sign showed 100% these cases. In none of the cases did reformatted
Right 6.62 6 0.588
sensitivity for left- and 83% sensitivity for right- images detect an injury unseen on axial images,
Left 5.5 3.5 0.201
sided diaphragmatic injuries in a recent investi- and in two cases the reformatted images were
gation of patients who sustained blunt trauma considered misleading.
more than 10 cm long [2, 7]. Gunshot wounds [2]. The sensitivity of this sign in our investiga- A recently described sign of diaphragmatic
can cause large blast injuries to the diaphragm, tion was only 52%. This lower sensitivity is per- injury, the thick crus sign, represents the subjec-
but stab wounds may result in smaller injuries haps related to the predominance of right-sided tive impression that the crus of the diaphragm is
[3, 16]. Therefore, helical CT might not detect injuries, which may be more difficult to diag- abnormally thickened [18]. We sought to estab-
penetrating diaphragmatic injuries as readily as nose on imaging [12], and to the relatively high lish a threshold measurement that could distin-
blunt injuries. However, our study found that the proportion of injuries caused by penetrating in- guish between normal and injured diaphragms
overall sensitivity and specificity of helical CT jury mechanisms in our study population. The to avoid the subjective nature of this sign. How-
for the detection of diaphragmatic injury were dependent viscera sign might require a large tear ever, the mean thickness of the diaphragmatic
84% and 77%, respectively. These results com- to become evident, and large tears may be more crura was not significantly greater for patients
pare favorably with the results of other CT inves- prevalent with blunt mechanisms of injury. Also, with an injured diaphragm than for those with
tigations that focused primarily on patients who on the left side, this sign may be observed when an uninjured diaphragm for the study popula-
sustained blunt trauma [3, 12]. Our favorable re- the tear occurs in the classic location for blunt tion as a whole or when segregating the popula-
sults, despite the fact that patients with penetrat- trauma: the posterolateral aspect of the dia- tion on the basis of blunt versus penetrating
ing injuries constituted a large portion of the phragm, between the spleen and abdominal injury mechanisms. It is possible that injured di-
study population, are at least in part due to the aorta [2]. This site is structurally weaker than the aphragms appear significantly thicker than un-
fact that injury trajectories seen on CT are useful remaining areas of the diaphragm [2, 3]. Given injured diaphragms on CT, but our study
markers of injuries caused by penetrating trauma that penetrating injuries do not occur preferen- population was too small to detect this differ-
[15, 16]. The trajectory of a penetrating injury tially at this site, this sign may be encountered ence. Additionally, because the thickness of the
may be visible on CT scans as localized soft-tis- less frequently in patients with penetrating dia- normal crus varies with patient age, the range of
sue swelling, subcutaneous emphysema, or focal phragmatic injury. Because this sign relies on the measurements for the normal diaphragmatic
extravasation of IV contrast material (Fig. 4). depiction of the cranial aspect of an abdominal crus would require analysis of a large number of
When injury to a solid or hollow organ is also viscus contacting posterior ribs, we found that patients representing a wide age range.
evident along this same path, diaphragmatic in- the most common cause of the absence of this The thick crus sign may be seen more fre-
jury is likely [15, 16, 27]. CT sign in patients with proven diaphragmatic quently in patients with a blunt injury, in
Although CT showed a moderately good injury was a large pleural effusion. Other investi- whom the site of injury is often more consis-
overall sensitivity of 84% for the detection of gators have also noted that pleural effusions may tent. Penetrating mechanisms cause injury on
diaphragmatic injury, the fact that sensitivity obscure diaphragmatic injury on CT [3, 10]. the basis of the type and trajectory of the pro-
was not higher is not unexpected. Studies of the Although the dependent viscera sign was jectile; therefore, a single measurement at one
natural aging of the diaphragm [29] and MR the most sensitive indicator of diaphragmatic portion of the diaphragm will probably be of
imaging of the diaphragm [30] have revealed injury in our series, the sensitivity of this sign little use. The subjective observation of dia-
that the entire diaphragm often cannot be com- was only 52%. Given that the overall sensitiv- phragm thickening as a manifestation of injury
pletely visualized with cross-sectional imaging. ity of helical CT for the detection of dia- may be more useful if extended to all portions
Couple this concept with the fact that penetrat- phragmatic injury was 84%, it is apparent of the diaphragm, not just the crus.
ing injuries are often smaller (and therefore that the various CT findings of diaphragm in- The two major limitations of our work are the
harder to detect) than blunt injuries and with the jury are often seen in isolation. This fact em- retrospective nature of the study and the rela-
limited specificity of several CT findings of dia- phasizes the idea that those who interpret CT tively small sample size. The latter is a problem
phragmatic injury, and it becomes clear why the scans of trauma patients must be familiar common to most studies about the utility of CT
CT diagnosis of traumatic diaphragmatic injury with all the CT findings of diaphragmatic in- for diagnosis of diaphragmatic injury and is asso-
is often a difficult one. jury and must be prepared to suggest the ciated with the relative rarity of preoperative CT
for an uncommon injury. For example, almost sagittal and coronal reconstructions are of lim- 13. Israel RS, McDaniel PA, Primack SL, Salmon CJ,
80% of the patients with proven diaphragmatic ited use in confirming or refuting this diagnosis. Fountain RL, Koslin DB. Diagnosis of diaphrag-
matic trauma with helical CT in a swine model. AJR
injury over the study period at our institution did The most sensitive sign for the detection of dia-
1996;167:637–641
not undergo CT before surgery; such a situation phragmatic injury was the recently described 14. Israel RS, Mayberry JC, Primack SL. Diaphragmatic
introduces selection bias in the review. dependent viscera sign, although observation of
Downloaded from www.ajronline.org by Washington State Univ. on 10/03/15 from IP address 134.121.47.100. Copyright ARRS. For personal use only; all rights reserved