| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Radiology, SS Annunziata Hospital, University of G. d'Annunzio,
Via P. Valignani 1, Chieti 66100, Italy.
2 Department of Radiology, Rm. M-391, University of California, 505 Parnassus
Ave., San Francisco, CA 94143-0628.
3 Department of Radiology, Thoracic Imaging Section, Rm. 1X 55A, Box 1325, San
Francisco General Hospital, 1001 Potrero Ave., San Francisco, CA 94110.
4 Centers for Disease Control and Prevention, National Center for Environmental
Health Centers for Disease Control and Prevention, MS E70, 1600 Clifton Rd.,
N.E., Atlanta, GA 30323.
Received January 4, 2002;
accepted after revision February 11, 2002.
Address correspondence to M. B. Gotway.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. The CT scans of 25 patients with surgically proven diaphragmatic injury and 22 patients with surgically confirmed uninjured diaphragms were blindly reviewed by five thoracic radiologists. Sagittal and coronal reconstructions were performed for 20 of the 25 patients with a proven diaphragmatic injury and for all the patients without a diaphragmatic injury. Scans were evaluated for findings suggestive of diaphragmatic injury and for associated injuries. Reviewers scored the usefulness of the reconstructed images for establishing the final diagnosis. Measurements of the right and left crura were performed to establish a threshold measurement that would enable radiologists to discriminate between a normal diaphragm and an injured diaphragm.
RESULTS. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of helical CT were 84%, 77%, 81%, 81%, and 83%, respectively. Scans showing active arterial extravasation of contrast material enabled reviewers to correctly identify diaphragmatic injury in two patients. Reconstructed images confirmed the correct diagnosis in three patients but supported an incorrect diagnosis in two. The mean thickness of the diaphragmatic crura (right and left) was not significantly greater in patients with an injured diaphragm than in those with an uninjured diaphragm.
CONCLUSION. Helical CT shows good sensitivity, specificity, and accuracy for the diagnosis of diaphragmatic injury. Coronal and sagittal reconstructions are of limited use in establishing or refuting this diagnosis. Active arterial extravasation of contrast material near the diaphragm should raise suspicion for injury. Crus measurements cannot be used to reliably distinguish between injured and uninjured diaphragms.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The normal diaphragm is a thin dome-shaped structure, a portion of which is oriented in the axial plane, which can make the imaging of diaphragmatic injury challenging [8]. Several studies have examined the use of conventional CT for the diagnosis of acute diaphragmatic injury [9,10,11]; a wide range of sensitivities and specificities have been reported. Only a few studies have evaluated the accuracy of single-detector helical CT for the diagnosis of diaphragmatic injury [2, 3, 12,13,14]. These studies focused primarily on the helical CT evaluation of patients with blunt trauma [2, 3, 12], and little data specifically examining the use of helical CT for diagnosing penetrating traumatic injury to the diaphragm exist.
Although traditionally many patients with penetrating injuries are taken directly to surgery without undergoing preoperative imaging, there is a growing trend toward more conservative treatment of such patients [15,16,17]. Therefore, CT is playing an increasingly important role in the assessment of penetrating thoracoabdominal injuries [15, 16]. Furthermore, although several reports [12, 14] have suggested that sagittal and coronal reconstructions have proven useful in confirming the diagnosis of diaphragmatic injury, little data are available that specifically evaluate the use of these techniques in trauma cases. The purposes of our investigation were to determine the accuracy of helical CT for the diagnosis of diaphragmatic injury in a patient population with a relatively high proportion of penetrating injuries, establish a threshold measurement of the crus that could be used to distinguish between an uninjured and an injured diaphragm, and describe an additional sign suggesting diaphragmatic injury: active arterial extravasation of contrast material at the level of the diaphragm.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The average age of the study population was 36.2 years (range, 17-82 years). Seventeen patients sustained right-sided injuries, and eight patients had left-sided injuries; there were no bilateral injuries. The mechanism of diaphragmatic injury was penetrating trauma (i.e., gunshot wound or stab wound) in 14 patients and blunt trauma (i.e., injury in a motor vehicle crash, struck by a car, or fall from a height) in 11 patients.
An additional 22 patients who sustained either blunt (n = 14) or penetrating (n = 8) trauma, had surgically confirmed intact diaphragms, and had undergone preoperative CT were randomly chosen and included as the control group, bringing the total study population to 47 patients.
CT Scans
All patients underwent helical CT (CTi; General Electric Medical Systems,
Milwaukee, WI) of the abdomen and pelvis after receiving water-soluble oral
contrast material (Gastrografin [meglumine diatrizoate]; Bracco Diagnostics,
Princeton, NJ) and IV contrast material (Omnipaque 300 [iohexol]; Nycomed,
Princeton, NJ). Scanning with a 7-mm collimation and a 7-mm reconstruction
interval was performed 70 sec after IV injection of contrast material at a
rate of 2 mL/sec. Imaging began at the lung bases, above the diaphragm, in all
patients.
Nine patients also underwent chest CT, which was performed before abdominal CT. CT scans of the chest were ordered at the discretion of the trauma service. Chest CT using a 5-mm collimation with a 3-mm reconstruction increment was performed 30 sec after IV injection of contrast material; the contrast material was administered at a rate of 3 mL/sec. Only one injection of contrast material was performed for patients who underwent both thoracic and abdominal CT examinations. One patient had dedicated 1-mm helical imaging through the lower chest and upper abdomen in addition to dedicated thoracic and abdominal CT. Chest CT extended from the cervicothoracic junction to the upper abdomen. Because the entire length of the diaphragm was not always completely covered by chest CT alone, axial and reformatted data from both chest and abdominal studies were presented for retrospective review. Images were photographed using soft-tissue window settings (level, 40 H; width, 440 H).
Digital CT data from 20 patients were available, thus allowing sagittal and coronal reconstructions to be performed. Reconstructed images were calculated using 0.7-mm spacing between images with commercially available software (Advantage Windows 4.0; General Electric Medical Systems). Approximately 40 reconstructed images were generated for each of the two planes.
CT Review
The CT studies from these patients were retrospectively reviewed by five
thoracic radiologists who were unaware of the surgical or autopsy findings.
The reviewers were aware only of the context of the study. Reviewers
specifically evaluated the imaging studies for an abnormally elevated
hemidiaphragm, diaphragmatic discontinuity, visceral herniation with or
without the collar sign, an abnormally thick-appearing crus, the
"dependent viscera" sign, and any other finding that could suggest
diaphragm injury. Sagittal and coronal reformations were reviewed
contemporaneously with axial images. Reviewers described each diaphragm as
injured, uninjured, or unknown and the reconstructed images for each case as
either helpful or not helpful in corroborating the diagnosis based on the
axial images. The reviewers understood that a diagnosis of
"unknown" would prompt surgical exploration to definitively
exclude diaphragmatic injury. The presence or absence of associated injuries
was also noted.
Diaphragm Measurements
For the entire study population, the thickness of the diaphragmatic crura
was measured in an attempt to establish a value that would enable radiologists
to distinguish between normal and injured diaphragms. The diaphragm crura were
measured by drawing perpendicular lines bisecting one another through the
center of the spinal canal at the level of L1 and then measuring the thickness
of the crus along a line drawn at a 45° angle to the intersection of these
two lines.
Statistical Analysis
The sensitivity, specificity, positive and negative predictive values, and
accuracy for the helical CT diagnosis of traumatic diaphragmatic injury were
calculated for the entire population, for patients with right- versus
left-sided injuries, and for those with a penetrating versus blunt mechanism
of injury. The sensitivity and specificity of the individual findings of
traumatic diaphragmatic injury were also calculated. The paired two-tailed
t test was used to compare the average thicknesses of the right and
left diaphragmatic crura of patients with an injured diaphragm and those of
patients with an uninjured diaphragm for the entire study population as well
as for patients with trauma from a blunt mechanism of injury and those with
trauma from a penetrating mechanism of injury.
Odds ratios were calculated for the ability of CT to detect diaphragmatic injuries caused by blunt and penetrating mechanisms of injury. The odds ratio represents the likelihood of CT detecting a diaphragmatic injury versus the likelihood of CT suggesting a diaphragmatic injury that was subsequently proven incorrect at surgery. These odds ratios were then compared with the Breslow and Day's test for homogeneity of odds ratios to determine whether CT performed significantly differently for the diagnosis of diaphragmatic injury in patients with a blunt versus penetrating mechanism of injury. Statistics were performed using commercially available software (SAS 8.2; Statistical Analysis System, Cary, NC).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The sensitivity, specificity, positive and negative predictive values, and accuracy for the detection of diaphragmatic injuries on helical CT were 84% (21/25), 77% (17/22), 81% (21/26), 81% (17/21), and 81% (38/47), respectively. The sensitivity and specificity of helical CT for penetrating injury were 86% (12/14) and 79% (11/14), respectively, and those for blunt injury were 82% (9/11) and 75% (6/8). The odds ratio for the CT detection of blunt diaphragmatic injury was 13.5 (CI, 1.47-123) and that for penetrating injury was 22.0 (CI, 3.1-157). Such large odds ratios imply a strong correlation between the CT diagnosis of traumatic diaphragmatic injury and surgical confirmation of diaphragmatic injury. The ability of CT to detect diaphragmatic injury on the basis of the mechanism—blunt or penetrating—did not differ significantly (p = 0.747).
The sensitivity and specificity in detecting diaphragmatic injuries in patients with a right- versus left-sided injury and in those with a penetrating versus blunt mechanism of injury as well as the individual helical CT findings of diaphragmatic injury are presented in Table 1. The most sensitive finding (52%) for the detection of diaphragmatic injury was the recently described dependent viscera sign [2] (Fig. 1). Direct visualization of diaphragmatic injury (Fig. 2) and the "thick crus" sign [18] (Fig. 3) were both 36% sensitive. The depiction of a missile or puncturing instrument trajectory was the most sensitive finding in patients with a penetrating diaphragmatic injury (Fig. 4). Active arterial extravasation of contrast material near the diaphragm (Fig. 2) suggested diaphragmatic injury in two patients and was the only sign of diaphragmatic injury in one of these two patients. For the entire population, the collar sign was the most specific (100%) manifestation of diaphragmatic injury (Fig. 5A,5B), although the trajectory of the injury in patients with penetrating injury was also 100% specific (Fig. 4).
|
|
|
|
|
|
|
False-negative interpretations (4/25) occurred most commonly in patients with other injuries abutting the injured diaphragm, such as a pleural effusion (Fig. 6) or perisplenic hematoma (Fig. 7).
|
|
Reviewers indicated that diagnostic confidence of a true-positive finding was increased by the sagittal and coronal reconstructions in three of the 20 true-positive cases (Fig. 5A,5B). Diaphragmatic injury was incorrectly suggested on reconstructed images in two cases (Fig. 8A,8B).
|
|
For the entire study population, the mean thickness of the diaphragmatic crura (right and left) was not significantly greater in patients with an injured diaphragm than in those with an uninjured diaphragm (Table 2). When classifying the cases according to mechanism of injury, we also found that the mean thickness of the diaphragmatic crura was not significantly greater in patients with an injured diaphragm than in those with an uninjured diaphragm (Table 2).
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CT has the major advantages of being widely available, rapid, noninvasive, and accurate for the assessment of the numerous injuries that are often associated with diaphragmatic injuries. Helical CT offers major advantages over conventional CT, particularly the ability to perform rapid volumetric acquisitions through the area of interest using a narrow collimation. Such techniques also allow high-quality sagittal and coronal reformations to be performed. These reformations have been shown to be beneficial in experimental models and in practice [3, 12,13,14].
Several studies have evaluated the utility of conventional and helical CT for the evaluation of diaphragmatic injury, with sensitivities and specificities ranging from 0% to 61% [2, 3, 9,10,11,12,13, 25, 26] and from 76% to 99% [2, 3, 10, 12, 13, 26], respectively. Several CT findings suggestive of diaphragmatic injury have been described, including apparent elevation of a hemidiaphragm, herniation of the abdominal contents into the thorax (with or without waist-like constriction of the herniated viscus, the socalled collar sign) [3], direct visualization of diaphragmatic discontinuity, an abnormally thickened diaphragmatic crus [18], evidence of penetrating injury with a trajectory projected to involve the diaphragm [15, 16, 27], and the recently described dependent viscera sign [2]. The latter manifests as the cranial aspects of the abdominal viscera (the liver on the right and stomach or bowel on the left) resting against posterior ribs because of the loss of diaphragmatic support after injury [2].
One of the purposes of our study was to examine the utility of helical CT for the detection of diaphragmatic injury in a patient population with a predominance of penetrating injuries (gunshot or stab wounds) as opposed to patients with blunt trauma injuries. Most CT studies of diaphragmatic injuries to date have focused on patients who sustained blunt trauma [3, 12], but these results may not be generalizable to patients with penetrating injuries. Mechanisms of diaphragmatic rupture from blunt trauma include rapid, large increases in intraabdominal pressure or shear stress [3, 6], whereas the injury mechanism from a penetrating injury is related to the missile or stabbing weapon itself. These differences in mechanism raise the possibility that the average extent of injury to the diaphragm may differ in patients who sustain blunt versus penetrating trauma [16, 28]. Many diaphragmatic injuries related to blunt trauma are large, often more than 10 cm long [2, 7]. Gunshot wounds can cause large blast injuries to the diaphragm, but stab wounds may result in smaller injuries [3, 16]. Therefore, helical CT might not detect penetrating diaphragmatic injuries as readily as blunt injuries. However, our study found that the overall sensitivity and specificity of helical CT for the detection of diaphragmatic injury were 84% and 77%, respectively. These results compare favorably with the results of other CT investigations that focused primarily on patients who sustained blunt trauma [3, 12]. Our favorable results, despite the fact that patients with penetrating injuries constituted a large portion of the study population, are at least in part due to the fact that injury trajectories seen on CT are useful markers of injuries caused by penetrating trauma [15, 16]. The trajectory of a penetrating injury may be visible on CT scans as localized soft-tissue swelling, subcutaneous emphysema, or focal extravasation of IV contrast material (Fig. 4). When injury to a solid or hollow organ is also evident along this same path, diaphragmatic injury is likely [15, 16, 27].
Although CT showed a moderately good overall sensitivity of 84% for the detection of diaphragmatic injury, the fact that sensitivity was not higher is not unexpected. Studies of the natural aging of the diaphragm [29] and MR imaging of the diaphragm [30] have revealed that the entire diaphragm often cannot be completely visualized with cross-sectional imaging. Couple this concept with the fact that penetrating injuries are often smaller (and therefore harder to detect) than blunt injuries and with the limited specificity of several CT findings of diaphragmatic injury, and it becomes clear why the CT diagnosis of traumatic diaphragmatic injury is often a difficult one.
Similar to Killeen et al. [12], we found that helical CT performed better for left-sided injuries (100% sensitivity) than for right-sided injuries (76% sensitivity). Other investigators have found that left-sided injuries are more readily detected on helical CT [3, 12], perhaps in part because of the relatively greater tissue contrast between fat in the left upper quadrant and the adjacent left hemidiaphragm than the poor tissue contrast between the right hemidiaphragm and the adjacent liver. These differences in tissue contrast may enhance direct visualization of diaphragmatic injuries.
The most sensitive sign of diaphragmatic injury in our study was the newly described dependent viscera sign. This sign showed 100% sensitivity for left- and 83% sensitivity for right-sided diaphragmatic injuries in a recent investigation of patients who sustained blunt trauma [2]. The sensitivity of this sign in our investigation was only 52%. This lower sensitivity is perhaps related to the predominance of right-sided injuries, which may be more difficult to diagnose on imaging [12], and to the relatively high proportion of injuries caused by penetrating injury mechanisms in our study population. The dependent viscera sign might require a large tear to become evident, and large tears may be more prevalent with blunt mechanisms of injury. Also, on the left side, this sign may be observed when the tear occurs in the classic location for blunt trauma: the posterolateral aspect of the diaphragm, between the spleen and abdominal aorta [2]. This site is structurally weaker than the remaining areas of the diaphragm [2, 3]. Given that penetrating injuries do not occur preferentially at this site, this sign may be encountered less frequently in patients with penetrating diaphragmatic injury. Because this sign relies on the depiction of the cranial aspect of an abdominal viscus contacting posterior ribs, we found that the most common cause of the absence of this CT sign in patients with proven diaphragmatic injury was a large pleural effusion. Other investigators have also noted that pleural effusions may obscure diaphragmatic injury on CT [3, 10].
Although the dependent viscera sign was the most sensitive indicator of diaphragmatic injury in our series, the sensitivity of this sign was only 52%. Given that the overall sensitivity of helical CT for the detection of diaphragmatic injury was 84%, it is apparent that the various CT findings of diaphragm injury are often seen in isolation. This fact emphasizes the idea that those who interpret CT scans of trauma patients must be familiar with all the CT findings of diaphragmatic injury and must be prepared to suggest the presence of an injury on the basis of a single CT finding.
We found sagittal and coronal reformations added little to the reviewers' ability to diagnose diaphragmatic injury. In contrast, Killeen et al. [12] found that reformatted images aided in the diagnosis of diaphragm injury, and experimental investigations reported that sagittal reformations were 92% sensitive for the diagnosis of diaphragm injury and were diagnostically superior to both coronal and axial reformations [13, 14]. In our series, reformatted images increased diagnostic confidence for only five (20%) of the 25 true-positive cases; the sagittal images were considered the most useful plane in three of these cases. In none of the cases did reformatted images detect an injury unseen on axial images, and in two cases the reformatted images were considered misleading.
A recently described sign of diaphragmatic injury, the thick crus sign, represents the subjective impression that the crus of the diaphragm is abnormally thickened [18]. We sought to establish a threshold measurement that could distinguish between normal and injured diaphragms to avoid the subjective nature of this sign. However, the mean thickness of the diaphragmatic crura was not significantly greater for patients with an injured diaphragm than for those with an uninjured diaphragm for the study population as a whole or when segregating the population on the basis of blunt versus penetrating injury mechanisms. It is possible that injured diaphragms appear significantly thicker than uninjured diaphragms on CT, but our study population was too small to detect this difference. Additionally, because the thickness of the normal crus varies with patient age, the range of measurements for the normal diaphragmatic crus would require analysis of a large number of patients representing a wide age range.
The thick crus sign may be seen more frequently in patients with a blunt injury, in whom the site of injury is often more consistent. Penetrating mechanisms cause injury on the basis of the type and trajectory of the projectile; therefore, a single measurement at one portion of the diaphragm will probably be of little use. The subjective observation of diaphragm thickening as a manifestation of injury may be more useful if extended to all portions of the diaphragm, not just the crus.
The two major limitations of our work are the retrospective nature of the study and the relatively small sample size. The latter is a problem common to most studies about the utility of CT for diagnosis of diaphragmatic injury and is associated with the relative rarity of preoperative CT for an uncommon injury. For example, almost 80% of the patients with proven diaphragmatic injury over the study period at our institution did not undergo CT before surgery; such a situation introduces selection bias in the review.
The sizes of the diaphragmatic injuries were not available from the surgical notes in most of the patients. Without knowing the sizes of the injuries, we cannot directly compare the performance of helical CT for the detection of blunt versus penetrating diaphragmatic injuries. Although our results compare favorably with those from studies performed primarily on patients who sustained blunt trauma, without the average size of the penetrating diaphragmatic injuries in our population, the role of helical CT in the diagnosis of penetrating diaphragmatic injury cannot be firmly established.
Our results might have been improved if a specific CT protocol had been implemented for imaging the diaphragm. Other investigators advocate CT evaluation of the region of the diaphragm with narrow collimation and reconstruction increments after routine CT evaluation for trauma to the abdomen [3]. These protocols increase spatial resolution and the quality of the reformatted images. With the increasing use of multidetector CT, rapid acquisition of narrowly collimated images may be possible without increased scanning time or increased heat-loading stress on the CT tube. Future investigations using this equipment may show improved results.
Our attempt to establish a crus measurement that would enable radiologists to distinguish between normal and injured diaphragms was limited by several factors. Our sample size was small; thus, our power to detect a difference between normal and injured diaphragms was limited. Because of the small size of our sample, we were not able to stratify measurements according to patient sex and age. Both of these factors are known to influence the thickness of the normal diaphragm. Finally, we performed our measurements at the level of the L1 vertebral body so that these measurements could be standardized across our population. However, the diaphragm has multiple muscular origins, so any measurement that attempts to differentiate normal from injured diaphragms will be useful only if the injury occurs at a site where the measurements were studied.
In conclusion, helical CT has good accuracy for the detection of diaphragmatic injury, but sagittal and coronal reconstructions are of limited use in confirming or refuting this diagnosis. The most sensitive sign for the detection of diaphragmatic injury was the recently described dependent viscera sign, although observation of the trajectory of missile or stab injuries in patients with penetrating trauma was also useful. The collar sign and the trajectory of the missile or stab injuries in patients with penetrating injury mechanisms were 100% specific for the detection of diaphragmatic injury. Because the sensitivity of any individual sign of diaphragmatic injury is not particularly high, familiarity with all CT findings of diaphragmatic injury is required, and the diagnosis must occasionally be made on the basis of a single finding.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  R. Kaewlai, L. L. Avery, A. V. Asrani, and R. A. Novelline Multidetector CT of Blunt Thoracic Trauma RadioGraphics, October 1, 2008; 28(6): 1555 - 1570. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. S. Restrepo, A. Eraso, D. Ocazionez, J. Lemos, S. Martinez, and D. F. Lemos The Diaphragmatic Crura and Retrocrural Space: Normal Imaging Appearance, Variants, and Pathologic Conditions RadioGraphics, September 1, 2008; 28(5): 1289 - 1305. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. P. Cantwell The Dependent Viscera Sign Radiology, February 1, 2006; 238(2): 752 - 753. [Full Text] [PDF]
|
|
|
|
 |
|
 A. Nchimi, D. Szapiro, B. Ghaye, V. Willems, J. Khamis, L. Haquet, C. Noukoua, and R. F. Dondelinger Helical CT of Blunt Diaphragmatic Rupture Am. J. Roentgenol., January 1, 2005; 184(1): 24 - 30. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Nishino, M. Kuroki, P. M. Boiselle, J. F. Copeland, V. Raptopoulos, and H. Hatabu Coronal Reformations of Volumetric Expiratory High-Resolution CT of the Lung Am. J. Roentgenol., April 1, 2004; 182(4): 979 - 982. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
