Article, Radiology

Diagnostic accuracy of the inverted grayscale rib series for detection of rib fracture in minor chest trauma

a b s t r a c t

Study objective: To assess whether inverted grayscale rib series, used alone or as an additional imaging modality, improves diagnostic accuracy of rib fractures of emergency medicine (EM) residents in minor chest trauma.

Methods: Twenty readers, including 5 junior and 5 senior EM residents and 10 fourth-year medical students, in- dependently reviewed 110 patients’ radiographs during 3 sessions. Session 1 used conventional grayscale rib se- ries, session 2 used inverted grayscale rib series, and session 3 used both conventional and inverted grayscale images. The McNemar test was used to compare the sensitivities and specificities of the diagnostic methods, and to compare their sensitivities and specificities for detecting more than 3 rib fractures. Interobserver agree- ment was assessed using Cohen ? analysis.

Results: For senior EM residents, there was no difference in sensitivity (P = .283) and accuracy (P = .888) be- tween conventional rib series and the double-modality method. For junior EM residents and medical students, the double modality offered higher diagnostic sensitivity (P b .001, P = .001) and accuracy (P = .006, P =

.002) than did conventional radiography. In cases with more than 3 rib fractures, who required specialist trauma care, the double modality provided greater sensitivity and accuracy among junior EM residents (P = .035 and P = .035, respectively) and medical students (P = .010, P = .010) than did conventional radiography.

Conclusions: In the diagnosis of rib fractures, the combined use of conventional and inverted grayscale rib series increases the diagnostic accuracy of less biased readers by conventional grayscale image like junior EM residents and medical students.

(C) 2015


Rib fracture and the associated pain in patients older than 45 years or with comorbidities like chronic obstructive pulmonary disease, cardi- ac disease, hepatic disease, renal disease, dementia, or coagulopathy have been linked to further complications and an increased mortality [1-3]. In addition, patients with 3 or more rib fractures need transfer to trauma center [4]. Currently, the best modality for diagnosing rib frac- tures and their Associated injuries is chest computed tomography, but it is relatively expensive, time-consuming, and not always available, and it exposes patients to significant radiation. Therefore, this diagnostic tool is especially inappropriate for use on emergency minor chest trauma

? ClinicalTrials ID: NCT02028611.

?? Contributions of authors: Dr Park managed all processes of recruiting subjects and ar- ranged radiologic data. Dr Cho took a part of statistic analysis. Dr Choi supervised whole study, especially to control radiologic review environment and to correct study design and statistic error.

* Corresponding author at: Emergency Department, College of medicine, Hanyang Uni- versity, 222, Wangsimni-ro(st), Seongdong-gu, Seoul 133-792, Korea. Tel.: +82 31 560

2055; fax: +82 31 560 2059.

E-mail address: [email protected] (H.J. Choi).

patients with no signs of intrathoracic complications. Ultrasound was found to offer better diagnostic sensitivity than plain x-ray in several studies; however, it sometimes failed to detect complications and was too dependent on operator experience [5-7]. However, single-plane x-ray often fails to detect rib fractures. In some studies, radiologists were found to misinterpret 40% of rib fractures as normal on chest radi- ography, and clinicians showed even lower sensitivity than radiologists [8,9].

In most emergency departments (EDs) in North America, radiolo- gists interpret radiographic images 24 hours a day, but in many coun- tries, including South Korea, this service is available only between 9 AM and 5 PM. Hence, emergency physicians may be required to review the images and decide on the course of treatment without the help of radiologists. To minimize emergency physicians’ mistakes, it is impor- tant to improve the quality of radiographic images. Digitization and the widespread use of Picture Archiving Communication System (PACS) have enabled image transformation, which may aid diagnosis. Some emergency physicians believe that grayscale inversion renders rib margins sharper than on conventional x-rays. However, the diagnos- tic use of inverted grayscale images in patients with rib fractures has not been tested. We hypothesized that the efficiency of detecting rib frac- ture by inverted grayscale image was varied across different levels of ra- diology experience. There are also no studies evaluating the influence of

0735-6757/(C) 2015

radiology experience on the ability to detect rib fractures using inverted grayscale rib series and chest posteroanterior (PA) view, with both oblique views.

The purpose of this study was to assess whether inverted grayscale rib series, used as an independent or an additional imaging modality, im- proved detection of rib fractures in minor chest trauma by emergency medicine (EM) residents according to their levels of radiology experience.


Study design and selection of participants

We performed a prospective observational study, using consecutive sample of the ED patients with rib fracture, from January 1, 2012, and May 31, 2013. The ED had about 35000 annual visits. Board-certified emergency physicians and EM residents have worked in this ED. This study was approved by the institutional review board of our medical center.

Two hundred fifty-three patients underwent rib series, and their fractures were confirmed by radiologists specializing in chest x-ray. We excluded 131 cases with prominent fractures detected by a first grade medical student and 67 cases with associated injuries including but not limited to hemothorax, pneumothorax, pulmonary contusion, and subcutaneous emphysema, which are more likely to imply rib frac- ture. Finally, 55 patients were enrolled.

Data collection and measurement

All rib series were performed using a digital radiographic device (TDR 4600-full; Geumsan Medical Company, Geumsan, South Korea). The im- aging consisted of a chest PA view, and left and right anterior oblique views. The oblique views enable the reader to examine parts of the tho- rax and ribs not visible on chest PA. For the oblique views, the patients were positioned at 45? to the radiographic device. If a patient could not stand, he or she was placed in a supine and rolling position at 45?.

The rib series were analyzed using a dedicated workstation (PiViewSTAR; Infinitt, Seoul, South Korea) and a 21-inch LCD monitor with a resolution of 2048 x 1536 pixels (IF2103MP, Wide, Gyeonggi- do, South Korea). The readers were 10 EM residents (5 junior: first or second year, and 5 senior: third or fourth year) and 10 medical students (fourth grade) who assessed the images independently during 3 sepa- rate reading sessions, from January to March 2014. During session 1, the readers analyzed conventional grayscale rib series (conventional method). During session 2, they analyzed inverted grayscale rib series (inverted method). During session 3, the conventional and inverted grayscale rib series were analyzed together (double-modality method) (Figure). To reduce recall and learning bias, the reading sessions were separated by 4-week intervals, and the order of sessions was random- ized. All the readers were asked not to perform inverted grayscale imag- ing during the study period. They were blinded to the patients’ clinical data, except for sex, age, and the site of pain (left or right side). Howev- er, they were aware that rib series had been clinically indicated in these patients due to chest trauma. During each reading session, the readers were asked to confirm the presence or absence of a rib fracture and identify and record the fractured rib (1st to 12th), and to grade their di- agnostic confidence in detecting the lesion on a 5-point scale. There was no time limit for the task, and the readers were allowed a 10-minute break every 30 minutes. All functions of the workstation were available: zooming, panning, and control of the width and level of the window. The standard of reference was a diagnosis made by an experienced chest radiologist.

Outcome measures

We defined diagnostic confidence of detection of rib fracture as a 5-point scale: 1 for “definitely no rib fracture,” 2 for “probably no rib

fracture,” 3 for “indeterminable,” 4 for “rib fracture probably present,” and 5 for “rib fracture definitely present.” The readers recorded each fractured rib and assigned their confidence score (3-5 points) for each suspected rib fracture on the answer sheet. When no rib fractures were suspected, a score of 1 or 2 was given.

We planned to perform subgroup analysis between senior and junior EM residents and, to effectively estimate the effect of radiolo- gy experience, added medical students as readers who had least ra- diology experience. The primary outcomes were sensitivity, specificity, and accuracy of detection of rib fracture according to dif- ferent grayscale methods and conventional method with inverted and double-modality respectively. We performed another subgroup analysis detecting cases with more than 3 rib fractures who clinically required specialist trauma care. In this subgroup analysis, sensitivity and accuracy were primary outcomes. We selected patients, matched for age and sex, as a control group. In primary outcomes, we statistically compared the conventional method with other 2 methods, respectively.

Data analysis

Sensitivity and specificity were compared between the conven- tional and other 2 methods using the McNemar test. This test was also used to compare the sensitivity and accuracy of the conven- tional and the other 2 methods in detecting cases with more than 3 rib fractures. Generalized estimating equation approach was used to compensate for any potential cluster effect of multiple lesions in the same patient. These statistical analyses were per- formed using commercially available software (SPSS 18.0; SPSS, Chicago, IL). Interobserver agreement among the 20 readers was assessed using Cohen ? statistic. P b .05 was considered to indicate statistical significance.


Characteristics of study subjects

In total, 139 rib fractures in 55 patients and 55 normal rib series were analyzed. Nine (16.4%) of 55 patients had 1 rib fracture, 21 (38.2%) had 2 fractures, and 25 (45.5%) had 3, 4, or 5 fractures. In 29 (52.7%) cases, the fractures were on the right side. There was no significant difference in patient characteristics between the groups with and without rib fractures (Table 1).

Detection of rib fractures based on grayscale imaging

For 20 readers, the double-modality method had higher diagnostic sensitivity (P b .001) than conventional radiography. For the 10 EM res- idents, the double modality also offered higher sensitivity (P = .011) than conventional imaging. In a subgroup analysis of the senior EM res- idents, the double modality offered no significant advantage over conventional radiography in terms of sensitivity (P = .283) or accuracy (P = .888). Conversely, among both junior EM residents and medical students, the double modality resulted in higher sensitivity (P b .001, P = .001) and accuracy (P = .006, P = .002). Senior EM residents achieved no statistical significant value (Table 2). Overall, there was fair to moderate agreement among the 20 readers in detecting rib fractures based on conventional (? = 0.400), inverted grayscale (? = 0.426), and both types of images (? = 0.435).

Detection of more than 3 rib-fractured patients

Of the enrolled patients, 25 had more than 3 rib fractures. Among all readers, among all residents, and among senior residents only, none of the diagnostic methods were significantly superior in terms of sensitiv- ity or accuracy in detecting more than 3 fractures. By contrast, for junior

Figure. Representative examples of rib fracture in conventional (A) and inverted grayscale (B) rib series images. Both images are left anterior oblique view in 78-year-old man and mag- nified fractured lesion in conventional (C) and inverted grayscale (D) images.

EM residents (P = .035 and P = .035, respectively) and medical stu- dents (P = .010 and P = .010) the double-modality method offered higher sensitivity and accuracy than conventional imaging. Specificity was 100.0% in all groups (Table 3).

Table 1

Patient demographic characteristics

Confirmed diagnosis

Rib fracture No rib fracture

No. of cases 55 55


Male (%) 34 (61.8) 33 (60.0)

Female (%) 21 (38.2) 22 (40.0)

Age (y)

Mean for all cases (SD) 56.1 (16.8) 56.4 (17.1)

Mean for men (SD) 51.4 (16.3) 52.4 (16.9)

Mean for women (SD) 63.8 (15.2) 62.3 (16.4) No. of fractures per case (%)


9 (16.4)


21 (38.1)


25 (45.5)

Side of fracture (%)


26 (47.3)


29 (52.7)


From the beginning, the PACS procedure for transforming conven- tional grayscale x-rays into inverted ones was embedded as a routine function in most PACS software. Inverted grayscale x-rays are easier to handle because they only require a simple click, so they remove the need for additional training on how to change conventional x-rays into inverted ones. Workers are accustomed to recognizing black letters or symbols on a white background. However, chest X-rays display white objects such as ribs or nodules on a black background, such as the lung. Therefore, in theory, the inversion of chest x-rays should be helpful to inexperienced readers in detecting lung parenchymal lesions.

In a conventional grayscale image, a rib fracture is seen as a gray le- sion consisting of a fracture line on the white background of the rib. In such cases, the inversion can make it easier to detect a fracture line in rib. Inversion has the additional effect of enhancing white or gray mate- rial on a dark background [10]. It has been reported that optical contrast adjustment yields better results when a bright object is set against a dark region [11]. On inverted grayscale images, the dark region is the bone, with a bright fracture line set against it. Some studies have sug- gested that, compared with conventional radiography, inverted gray- scale imaging improves detection of bone or lung lesions [12,13]. A prior study reported that inverted grayscale imaging in dental practice enhanced the contours of facial bone fractures, improving their

Table 2

Characterization of rib fractures by readers according to different diagnostic methods

Detection of rib fracture

Reader group

Grayscale imaging method

Conventional Inverted Pa Both types Pb

All readers (n = 20)


59.9% (1666/2780)

60.6% (1684/2780)


65.0% (1806/2780)



85.2% (937/1100)

85.5% (940/1100)


86.1% (947/1100)



67.1% (2603/3880)

67.6% (2624/3880)


71.0% (2753/3880)


All EM residents (n = 10) Sensitivity

61.3% (852/1390)

60.4% (840/1390)


65.4% (909/1390)



89.3% (491/550)

88.9% (489/550)


87.6% (482/550)



69.2% (1342/1940)

68.5% (1328/1940)


71.7% (1391/1940)


Senior residents only (n = 5)

Sensitivity 64.9% (451/695)

60.9% (423/695)


61.6% (428/695)



93.8% (258/275)

92.4% (254/275)


97.5% (268/275)



73.1% (709/970)

69.8% (677/970)


71.8% (696/970)


junior residents only (n = 5) Sensitivity

57.7% (401/695)

59.9% (416/695)


69.2% (481/695)



84.7% (233/275)

85.5% (235/275)


77.8% (214/275)



65.4% (634/970)

67.1% (651/970)


71.7% (695/970)


All students (n = 10) Sensitivity

58.6% (814/1390)

60.8% (845/1390)


64.5% (897/1390)



81.1% (446/550)

82.0% (451/550)


84.6% (465/550)



65.0% (1260/1940)

66.8% (1296/1940)


70.2% (1362/1940)


P value by generalized estimating equation method.

Data from 55 cases with 139 rib fractures and other 55 cases without rib fracture.

a P value in the McNemar test, conventional vs inverted grayscale images.

b P value in the McNemar test, conventional vs both types together.

detection [14]. On this basis, we hypothesized that inverted grayscale rib series could improve the detectability of rib fractures. Therefore, if a rib fracture is indistinct on a conventional chest x-ray, reviewing it on an inverted grayscale chest x-ray should provide better discrimina- tion. Because inverted grayscale x-rays have a more familiar and intui- tive appearance than conventional ones, their use should also improve the diagnostic power of other kinds of x-ray. It would be useful to teach some individuals without previous experience of conventional x-rays, such as premedical students, how to use both conventional and inverted grayscale x-rays, so that we would be able to compare the discriminating powers of the 2 grayscale settings without any

experience bias. If, as a result of such a comparison, inverted grayscale x-rays were included in the curriculum of medical school, the decisions of clinicians such as emergency physicians would be improved.

According to published reports, inverted images may offer some ad- vantage only to radiology residents in detecting pulmonary nodules, and other published reports showed that there was limited benefit only in dental practice and facial bone fractures [12,15,16]. Moreover, many studies contradicted the advantage of inverted chest x-ray, reporting that radiologists could not detect lung nodules effectively using grayscale inversion [17-19]. A recent study found that inverted grayscale did offer a diagnostic advantage in detecting lung nodules,

Table 3

Detection of more than 3 rib-fractured patients

Reader group

All readers (n = 20)

Detection of >=3 rib fractures Grayscale imaging method

Conventional Inverted Pa Both types Pb


44.4% (222/500)

44.8% (224/500)


51.6% (258/500)



100.0% (500/500)

100.0% (500/500)

100.0% (500,500)


72.2% (722/1000)

72.4% (724/1000)


75.8% (758/1000)


All EM residents (n = 10) Sensitivity

49.2% (123/250)

45.2% (113/250)


54.0% (135/250)



100.0% (250/250)

100.0% (250/250)

100.0% (250/250)


74.6% (373/500)

72.6% (363/500)


77.0% (385/500)


Senior Residents only (n = 5)

Sensitivity 54.4% (68/125)

49.6% (62/125)


52.8% (66/125)



100.0% (125/125)

100.0% (125/125)

100.0% (125/125)


77.2% (193/250)

74.8% (187/250)


76.4% (191/250)


Junior residents only (n = 5) Sensitivity

44.0% (55/125)

40.8% (51/125)


55.2% (69/125)



100.0% (125/125)

100.0% (125/125)

100.0% (125/125)


72.0% (180/250)

70.4% (176/250)


77.6% (194/250)


All students (n = 10) Sensitivity

39.6% (99/250)

44.4% (111/250)


49.2% (123/250)



100.0% (250/250)

100.0% (250/250)

100.0% (250/250)


69.8% (349/500)

72.2% (361/500)


74.6% (373/500)


Data from 25 cases with more than 3 rib fractures and other 25 cases without rib fracture.

a P value in the McNemar test, conventional vs inverted grayscale images.

b P value in the McNemar test, conventional vs both types together.

but only in limited cases [20]. The study used a particular methodology, the Dorfman-Berbaum-Metz Multiple-Reader, Multiple-Case, for re- ceiver operating characteristic analysis of multiple readers in multiple cases. The method examined 3 types of situation: (1) random readers and random cases, (2) random readers and fixed cases, and (3) fixed readers and random cases. The study showed that the diagnostic perfor- mance of inverted grayscale chest x-rays was superior in fixed cases, but that their superiority was not reproducible in other cases. Most of the studies with a focus on dental practice and facial bone fractures had only small numbers of patients, or they only assessed interobserver agreement, so their findings could not be generalized to other lesions [14,16].

No study to date has investigated the diagnostic advantage of inverted grayscale imaging over conventional x-ray in detecting rib fractures, or its dependence on physicians’ radiology experience. In our study, we classified the reviewers into 3 groups: those with more than 3 years of radiology experience, with less than 2 years of experi- ence, and with no experience. We found that inverted grayscale imag- ing, either as a standalone method or in combination with conventional x-rays, offered no diagnostic benefits to the physicians with the most experience. This should not be interpreted as a lack of any diagnostic benefit, but as a confirmation that extensive experience of conventional radiography leads to more accurate diagnosis and cre- ates an inevitable bias in favor of the conventional method. This bias could only be avoided by training Novice doctors to use both methods from the beginning: only then we could estimate the diagnostic power of each method independently. Therefore, although this study could not prove the efficacy of inverted grayscale image in radiology- experienced emergency physicians, our results suggest that less biased readers like junior EM residents should be encouraged to use the double-modality method from the beginning of their training, so that they may maintain higher rates in detecting rib fracture by the double-modality method than when solely using the conventional method.

It was surprising that in senior EM residents, the sensitivity and ac- curacy of the double-modality method appeared to be lower than those of conventional imaging, because the double-modality method in- cluded the same conventional images. When questioned after 3 sessions of chest compression, more than half the readers reported using the conventional images first to make a preliminary diagnosis and the inverted images second to confirm the initial decision. Because the se- nior residents were more accustomed to conventional x-ray, their Initial diagnosis was seldom modified by reviewing the inverted images. The readers also reported that all 3 reviewing sessions were of similar length, which suggests that the time taken to examine the conventional x-ray images was shorter during session 3 (double modality) than ses- sion 1 (conventional images only). Not taking sufficient time to carefully review both sets of images may partly explain the lower sensitivity of the double modality.

Patients with 3 or more rib fractures and older than 14 years are more likely to require trauma care at a tertiary center due to the in- creased risk of spleen injury, liver injury, hemothorax, and pneumotho- rax [4]. Our findings suggest that the double-modality method may improve the proportion of correct decisions made for these patients by junior residents and medical students. This is particularly important when trauma or emergency physicians without sufficient radiology ex- perience have to decide whether specialist trauma care is required.

This study has a few limitations. Although our focus was whether inverted grayscale image had any benefit in detecting rib fracture com- pared with the conventional method, according to levels of radiology experience in EM residents, we included fourth-year medical students as additional readers instead of trainee radiologists or other physicians. We hypothesized that physicians with the least experience of conven- tional imaging would benefit the most from inverted imaging. Thus,

including EM residents and medical students in our study was a way of minimizing the experience bias toward conventional x-ray. Another potential issue with the study design is that the readers reviewed the same cases 3 times. Despite randomization of the sessions and the 4- week intervals between each session, it is still possible that recall bias affected the readings. One more potential criticism is the inclusion of pa- tients with more than one rib fracture, because one prominent rib frac- ture could reasonably increase the expectation of another. Unfortunately, the number of patients with only one rib fracture avail- able for inclusion was very small, which is not representative of real clinical situations and could be seen as a selection bias. To compensate this clustering bias, we used specific statistic method, generalized esti- mating equation.

In summary, inverted grayscale rib series did not improve the rate of detection of rib fractures over conventional x-rays regardless levels of radiology experience, but using both sets of images at the same time im- proved diagnostic sensitivity and accuracy among less biased readers by conventional image like junior EM residents or medical students.


This work was partially supported by Soonchunhyang University Re- search Fund and Hanyang University (HY-2013-MC) Research Fund.


  1. Holcomb JB, McMullin NR, Kozar RA, Lygas MH, Moore FA. Morbidity from rib frac- tures increases after age 45. J Am Coll Surg 2003;196:549-55.
  2. Bergeron E, Lavoie A, Clas D, Moore L, Ratte S, Tetreault S, et al. Elderly trauma pa- tients with rib fractures are at greater risk of death and pneumonia. J Trauma 2003;54:478-85.
  3. Byun JH, Kim HY. Factors affecting pneumonia occurring to patients with multiple rib fractures. Korean J Thorac Cardiovasc Surg 2013;46:130-4.
  4. Lee RB, Bass SM, Morris Jr JA, MacKenzie EJ. Three or more rib fractures as an indicator for transfer to a level I trauma center: a population-based study. J Trauma 1990;30: 689-94.
  5. Chan SS. Emergency bedside ultrasound for the diagnosis of rib fractures. Am J Emerg Med 2009;27:617-20.
  6. Hurley ME, Keye GD, Hamilton S. Is ultrasound really helpful in the detection of rib fractures? Injury 2004;35:562-6.
  7. Kara M, Dikmen E, Erdal HH, Simsir I, Kara SA. Disclosure of unnoticed rib fractures with the use of ultrasonography in minor blunt chest trauma. Eur J Cardiothorac Surg 2003; 24:608-13.
  8. Aukema TS, Beenen LF, Hietbrink F, Leenen LP. Initial assessment of chest x-ray in Thoracic trauma patients: awareness of specific injuries. World J Radiol 2012;4: 48-52.
  9. Griffith JF, Rainer TH, Ching AS, Law KL, Cocks RA, Metreweli C, et al. Sonography compared with radiography in revealing acute rib fracture. AJR Am J Roentgenol 1999;173:1603-9.
  10. Jain AK. Fundamentals of digital image processing. Englewood Cliffs, NJ: Prentice Hall; 1989.
  11. Westheimer G. visual acuity with reversed-contrast charts: I. Theoretical and psy- chophysical investigations. Optom Vis Sci 2003;80:745-8.
  12. Sheline ME, Brikman I, Epstein DM, Mezrich JL, Kundel HL, Arenson RL, et al. The di- agnosis of pulmonary nodules: comparison between standard and inverse digitized images and conventional chest radiographs. AJR Am J Roentgenol 1989;152:261-3.
  13. van der Stelt PF. Better imaging: the advantages of digital radiography. J Am Dent Assoc 2008;139:7S-13S [Suppl.].
  14. van der Stelt PF. Filmless imaging: the uses of digital radiography in dental practice. J Am Dent Assoc 2005;136:1379-87.
  15. Sakakura CE, Loffredo Lde C, Scaf G. Diagnostic agreement of conventional and inverted scanned panoramic radiographs in the detection of the mandibular canal and the mental foramen. J Oral Implantol 2004;30:2-6.
  16. Punhani N, Daniel J. Conventional radiography and inverted digitized imaging in the detection of maxilloFacial fractures. J Dent 2011;1:63-7.
  17. De Boo DW, Uffmann M, Bipat S, Boorsma EF, Scheerder MJ, Weber M, et al. Gray- scale reversal for the detection of pulmonary nodules on a PACS workstation. AJR Am J Roentgenol 2011;197:1096-100.
  18. Lungren MP, Samei E, Barnhart H, McAdams HP, Leder RA, Christensen JD, et al. Gray-scale inversion radiographic display for the detection of pulmonary nodules on chest radiographs. Clin Imaging 2012;36:515-21.
  19. Kheddache S, Mansson LG, Angelhed JE, Denbratt L, Gottfridson B, Schlossman D, et al. Digital chest radiography: should images be presented in negative or positive mode? Eur J Radiol 1991;13:151-5.
  20. Robinson JW, Ryan JT, McEntee MF, Lewis SJ, Evanoff MG, Rainford LA, et al. Grey- scale inversion improves detection of lung nodules. Br J Radiol 2013;86:20110812.

Leave a Reply

Your email address will not be published. Required fields are marked *