Digital radiography in the diagnosis of toddler's fracture. (Original Article).

Background: The objective of this study was to evaluate the accuracy of digital imaging in the diagnosis of toddler's fractures.

Methods: Medical records for a 9.4-year period were reviewed to locate children whose initial radiographs were interpreted as normal by a pediatric radiologist and whose subsequent bone scans or follow-up radiographs showed toddler's fractures. Radiographs from these children (ie, positive controls) and from children without toddler's fractures (ie, negative controls) were digitized to create a film bank that was reviewed by a panel of 14 physicians with various medical backgrounds. Medical records were reviewed for demographic information, findings on history and physical examination, and radiographic and laboratory tests.

Results: Pediatric radiology physicians correctly diagnosed 73.2 +/- 5.4% of the digitized images, as compared with pediatric emergency physicians, 66.7 [+ or -] 6.5% and residents/fellows, 57.1 [+ or -] 6.9%.

Conclusion: Digitized images may be helpful in evaluating limping children with suspected toddler's fractures, possibly eliminating the need for further diagnostic studies.

**********

Digital imaging and teleradiology are becoming standard practice, often replacing conventional hard-copy screen-film radiographic interpretation. Radiologists are now able to receive and interpret images from multiple hospitals at a central location, and smaller rural hospitals are able to obtain timely radiographic consultation by digital transmission. Currently, numerous radiology departments are implementing or assessing conversion to a digital "filmiess" radiology department. Previous studies have compared digitized radiographs and plain film interpretation with mixed results. (1-4) Digital imaging has many benefits over conventional plain film radiography, including archiving, smaller storage requirements, expedient acquisition of both the current and comparison images, viewing from remote sites with radiologic consultation, and decreased likelihood of lost films. In addition, digital imaging may allow for manipulation of images to compensate for suboptimal exposure and thus improve diagnostic accuracy in. some cases.

One possible application for digital imaging is in the diagnosis of occult pediatric fractures. Dunbar et [al.sub.5] first described a group of subtle childhood tibial fractures termed toddler's fractures in 1964. These fractures were defined as subtle, nondisplaced, oblique fractures occurring in the distal tibial shaft in children 9 months to 3 years of age. Toddler's fractures are frequently undetected on initial radiographs but may be evident on follow-up radiograplis after some degree of healing has occurred. Nuclear medicine bone scan imaging may also be used to make the diagnosis. If the initial diagnosis of tibial fracture cannot be made, frequently other more serious conditions such as a septic joint or osteomyelitis are considered. Numerous laboratory and radiographic tests may be obtained, resulting in a substantial cost to the patient. The objective of this study was to evaluate the accuracy of digitally processed images of conventional plain radiographs in detecting subtle tibial fractures not d iagnosed on initial radiographic interpretation (Fig. 1).

Methods

The human research advisory committee of our institution approved the study, and informed consent was not required. The medical records of children from 0 to 4 years of age with a diagnosis of tibial fracture (ICD code 823.80) between July 1, 1988, and November 30, 1998, were reviewed, and 172 charts were identified. Charts were excluded for patients who were under evaluation for child maltreatment syndrome, had obvious tibial fractures diagnosed on initial radiographs, or had tibial fractures as a result of metabolic bone disease. A total of 13 patients of the original 172 were identified as meeting the criteria. Age-matched controls were identified using plain film radiograplis of the opposite extremity obtained for comparison view or randomly identified tibial radiographs interpreted as normal and without subsequent diagnosis of a fracture.

Patient medical records were reviewed for demographic information, clinical history, physical examination findings, number and type of radiographs obtained, and laboratory or other diagnostic tests. Data were tabulated using Excel software (Microsoft Corp., Redmond, WA) and analyzed with standard descriptive statistics. Sensitivity, specificity, and accuracy were calculated for individual reviewers, for the reviewer panel as a whole, and for professional subgroups. A k statistic was used to describe the variability within professional subgroups. Data are presented as mean [+ or -] standard deviation.

A total of 14 reviewers with various medical backgrounds were selected to interpret the digital film images. Reviewers included 7 board-certified pediatric radiologists, 3 board-certified pediatric emergency medicine physicians, 2 senior emergency medicine residents, 1 pediatric radiology fellow, and i senior radiology resident. Reviewers were grouped by the following three professional subgroups: pediatric radiologists (n = 7), pediatric emergency medicine physicians (a = 3), and residents/fellows (n = 4).

Reviewers were instructed on the inclusion and exclusion criteria of the study and were given information for using the digital viewing terminal. Reviewers were informed that the digitized images consisted of both positive controls (ie, digitized images from patients with proven toddler's fractures previously read as "normal" by a pediatric radiologist) and negative controls (ie, digitized images obtained from radiographs that were obtained as "comparison views or from randomly selected radiographs, without subsequent diagnosis of toddler's fracture). Reviewers were not provided the number of either positive or negative controls in the film bank and were not provided with patient histories. Two references were made available to reviewers that included the original description of toddler's fracture and a recently described subtle proximal metaphyseal tibial fracture identified by Tschoepe et al. (6)

The digitized film bank consisted of 32 sets of films. The sets of films were equally composed of 16 sets of true-positive films (ie, toddler's fractures that were confirmed by bone scan or follow-up radiographs revealing a healing fracture line) and 16 sets of films from normal age-matched controls selected by the process described. The 16 sets of true-positive films were obtained from 13 patients. One of the patients had two separate fractures (a proximal and a distal tibial fracture) more than a year apart. Two of the patients had two sets of films each, obtained on different days throughout the evaluation process, all of which were read by pediatric radiologists as normal.

The radiographs in the test set were digitized to a 2,048 X 2,048 x 12-bit digital matrix with a Lumisys model 150 laser film digitizer (Lumisys, Sunnyvale, CA). The viewing monitor was a Megascan monitor DG-4820P (Megascan, West-ford, MA) with 1726 X 2304 X 8-bit resolution. The variable spot size was 50 [micro]m with 2048 pixels over 8 to 14 inches. The density resolution and the density range were 0.001 and 0 to 3.5 optical density, respectively.

Results

Radiographs were identified from 13 patients who met the study criteria. That is, 13 patients had tibial radiographs that were read as normal by pediatric radiologists and subsequently were found to have toddler's fractures when either bone scan or follow-up radiographs revealed a healing fracture. Study subjects were 8 to 48 months of age with a mean age of 20.9 [+ or -] 10.8 months. Eight patients were female.

Accuracy among the pediatric radiologist subgroup was higher 73 2 [+ or -] 5.4%, than among the other two subgroups of pediatric emergency medicine 66.7 [+ or -] 6.5%, and residents/fellows 57.1 [+ or -] 6.9%. The accuracy for the reviewer group as a whole was 67.2 [+ or -] 9.1%. The sensitivity of digital imaging for the reviewer panel by subgroups was as follows: pediatric radiologists, 65.2%; pediatric emergency medicine physicians, 52.1%; and residents/fellows, 62.5%. The specificity for the reviewer panel by subgroups was as follows: pediatric radiologists, 79.5%; pediatric emergency medicine physicians, 81.3%; and residents/fellows, 53.2%.

The agreement of the 14 reviewers was measured using the generalized k statistic. The agreement within the individual subgroups of pediatric radiologists and pediatric emergency medicine physicians was similar and suggested fair interrater reliability (k = 0.43 8 and 0.408, respectively), whereas agreement within the resident/fellow group was considerably lower (k = 0.087). The most common chief complaint for study subjects (n = 14, including 2 fractures separated by more than a year in one subject) was the inability to bear weight on the affected extremity (n = 10), followed by limping (n = 3) and pain and swelling (n = 1). Predominant findings on physical examination were local tenderness (n = 4), no abnormality (n = 4), pain with range of motion (n = 3), local swelling (n = 2), and local warmth (n = 1). Multiple physician encounters, including subspecialty consultation, were common for the study population (Table 1). The average time to radiographic diagnosis of toddler's fracture from initial presentatio n to our facility was 2.5 [+ or -] 2.4 days. Two patients were hospitalized for 2 and 4 days each. Laboratory and radiographic tests performed for the entire study population are summarized in Table 2. Cumulative patient charges for these tests were $11,216.

Discussion

Although tibial fractures are common fractures in children beginning to walk, the diagnosis may be difficult as a history of trauma is often trivial. (7) The physical examination may show few positive findings and may be complicated by a lack of cooperation and the inability of children in this age group to verbalize complaints. (7) Early, accurate diagnosis of these fractures is important to provide reassurance to families and treating physicians and to avoid unneeded, expensive diagnostic tests. A recent study (8) suggested that a "second look" at tibial radiographs of toddlers may reveal the subtle fracture missed on initial radiographic interpretation.

This study was conducted to evaluate whether digitized images are beneficial in making the diagnosis of toddler's fractures when films were initially read as normal by pediatric radiologists. The reviewers in this study, as a whole, correctly diagnosed 67.2% of toddler's fractures and controls using digitized images, and the pediatric radiologists had the highest accuracy within the three professional subgroups. These data suggest that digitized imaging may be a helpful diagnostic aid in the initial evaluation of the child with a suspected toddler's fracture when initial radiographs are considered normal. Although the sensitivity and specificity for digital imaging in our study were relatively low, the films selected for digital manipulation were from a population in which the diagnosis originally eluded the examining physicians and pediatric radiologists, leading to multiple, costly diagnostic and laboratory tests. It is possible that digital manipulation of "normal" tibial radiographs in toddlers with limp ing or refusal to bear weight may lead to a more timely diagnosis and the possible avoidance of subsequent expensive diagnostic tests.

Since the original description, (5) the term toddler's fracture has been applied loosely to other subtle tibial fractures, subtle hindfoot fractures, and tibial fractures in children older than 3 years of age. (8,9) The 13 patients in our study had a total of 14 fractures (1 patient having 2 separate fractures during the course of a year). Six fractures were distal in location, consistent with the original description of toddler's fracture. An additional 6 fractures (43%) were proximal tibial fractures, suggesting that this type of fracture may be more prevalent than previously thought (Figs. 2 and 3). (6) There were also 2 midshaft fractures. Previous studies have compared the diagnostic accuracy of digital imaging and plain film radiography with varying results. (1-4) Weaknesses of these studies include lower resolution of the images, unfamiliarity with the imaging system, and reviewers with minimal experience in reading digitized images. The progression of technology since the publication of these initial studies has led to improved resolution of digitized images.

A recent study by Youmans et a1 (10) concluded that digital images were inferior to plain films for skeletal survey in cases of suspected child abuse. This study, although using a 2k X 2k resolution digitizer and experienced reviewers, resulted in an underestimation of the fractures. However, these fractures were identified retrospectively after image manipulation on the monitor resulting in optimal contrast adjustment and panning of the entire image. Murphey et al (11) evaluated digital images of nondisplaced fractures and concluded that resolutions coarser than 2k X 2k would result in a significant number of fractures going undetected. Because the images used in this study were digitized from plain films, poor image resolution may have had an impact on the study's results.

Cox et al (12) compared plain film, digital hard copies at 2k X 2k resolution and an interactive digital monitor with 2k X 2k resolution of chest radiographs containing common radiographic abnormalities. Their results indicated that the digital hard copy was equivalent to or better than plain films in detecting the abnormalities, whereas the interactive display system was inferior in some instances. These data suggest that the diagnostic accuracy was influenced by reviewer familiarity with the monitor system and adjustment parameters.

Thus, it appears that identification of fractures using digitized images may be dependent on reviewer experience in reading digital images and reviewer familiarity with the resolution parameters of the monitor. Increased experience with digital image manipulation most likely will result in optimal accuracy. Our study used a 2k X 2k resolution digitizer and reviewers with a variety of experience ranging from emergency medicine residents to board-certified pediatric radiologists'. The majority of our reviewers (9/14) had previous experience manipulating and interpreting digitized images, whereas the remainder (5/14) had little or no experience.

Potential weaknesses of our study include a lack of comparison of the digital images to their plain film counterpart and possible reviewer bias. The reviewer panel reviewed only the digitized images and not the plain film radiographic counterparts; thus, no direct comparison between these two modalities was made. However, individuals from the pediatric radiology subgroup had previously interpreted all plain radiographs as normal at the time that radiographic imaging was done.

Although three of us served on the reviewer panel, each author's accuracy for the diagnosis of toddler's fracture fell within accuracy range of the other members of his or her respective professional subgroup. One of us (J.R.B.) assembled the film bank and did not serve on the reviewer panel. In addition, the reviewer panel as a whole may have been biased in that they received a brief tutorial on toddler's fractures before reviewing the digitized film bank. However, pediatric radiologists reviewing radiographs of non-weight-bearing children in the toddler age group routinely consider toddler's fracture in the differential diagnosis.

Our study suggests that digitized images may play a helpful role in the diagnosis of toddler's fractures. Based on the sensitivity of digital imaging from this study, this modality may detect up to 65% of toddler's fractures that were initially undiagnosed on plain radiographs. The use of digital imaging in the setting of a limping toddler with "normal" tibial radiographs may help to eliminate further diagnostic tests that may be more costly and time consuming. Future prospective studies using digital imaging should be conducted to further define the role of this technology in evaluating the limping toddler.

Table 1

Physician encounters with 13 patients with toddler's fractures

Type of consultation             No. of visits

Physician encounters
 Community physicians                  7
 General pediatric clinic              2
 Pediatric emergency department        6
 Hospitalization                       2
Subspecialty consultations
 Orthopedics                          14
 Infectious disease                    1

Table 2

Diagnostic tests performed in 13 patients with toddler's fractures

Test                        No.

Standard radiography        49
Nuclear medicine bone scan   8
Ultrasonography              4
Computed tomography          1
Laboratory studies (a)      22

Total                       84

(a)Laboratory studies included complete blood count, erthrocyte
sedimantation rate, urinalysis, and blood culture.

Acknowledgments

We thank the physicians of the Departments of Radiology and Pediatrics (Section of Emergency Medicine) of the University of Arkansas for Medical Sciences and Arkansas Children's Hospital for their contributions to this study.

Accepted June 18, 2002.

References

(1.) Balter S. Fundamental properties of digital images. Radiographics 1993; 13:129-141.

(2.) Wegryn SA, Piraino DW, Richmond BJ, Schluchter MD, Uetani M, Freed HA, et al. Comparison of digital and conventional musculoskeletal radiography: An observer performance study. Radiology 199C;175:225-228.

(3.) Wilson AJ, Hodge JC. Digitized radiographs in skeletal trauma: A performance comparison between a digital workstation and the original film images. Radiology 1995;196:565-568.

(4.) Scott WW Jr, Rosenbaum JE, Ackerman SJ, Reichle RL, Magid D, Weller JC, et al. Subtle orthopedic fractures: Teleradiology workstation versus film interpretation. Radiology 1993;187:811-815.

(5.) Dunbar JS, Owen HF, Nogrady MB, McLeese R. Obscure tibial fracture of infants: The toddler's fracture. J Can Assoc Radiol 1964;15:136-144.

(6.) Tschoepe EJ, John SD, Swisehuk LE. Tibial fractures in infants and children: Emphasis on subtle injuries. Emerg Radio! 1998;5:245-252.

(7.) Tenenbein M, Reed MH, Black GB. The toddler's fracture revisited. An, J Emerg Med 1990;8:208-211.

(8.) John SD, Moorthy CS, Swisehuk LE. Expanding the concept of toddler's fracture. Radiographics 1997;17:367-376.

(9.) Oudjhane K, Newman B, Oh KS, Young LW, Girdany BR. Occult fractures in preschool children. J Trauma 1988;28:858-860.

(10.) Youmans DC, Don S, Hildebolt C, Shackelford GD, Luker GD, McAlister WH. Skeletal surveys for child abuse: Comparison of interpretation using digitized images and screen-film radiographs. AJR Am, J Roentgenol 1998;171:1415-1419.

(11.) Murphey MD, Bramble JM, Cook LT, Martin NL, Dwyer SJ III. Non-displaced fractures: Spatial resolution requirements for detection with digital skeletal imaging. Radiology 1990; 174:865-870.

(12.) Cox GG, Cook LT, MeMillan JH, Rosenthal SJ, Dwyer SJ III. Chest radiography: Comparison of high-resolution digital displays with conventional and digital film. Radiology 1990;176:771-776.

RELATED ARTICLE: Key Points

* The toddler's fracture as originally described is a subtle fracture of the distal tibia in children 9 months to 3 years of age.

* Toddler's fractures are frequently undetected on initial radiographs, and other diagnostic tests may be performed as a result.

* More recently, a subtle proximal initial metaphyseal toddler's fracture has been reported.

* Digitized images may complement the diagnostic accuracy of plain radiographs in toddler's fractures.

From the Department of Emergency Medicine; Department of Pediatrics, Sections of Pediatric Emergency Medicine and Clinical Pharmacology and Toxicology; and Department of Radiology, University of Arkansas for Medical Sciences and Arkansas Children's Hospital, Little Rock.

Reprint requests to Laura P. James, MD, Clinical Pharmacology and Toxicology, Pediatric Emergency Medicine, Arkansas Children's Hospital, 800 Marshall Street, Little Rock, AR 72202. Email: [e-mail]

Copyright [C] 2003 by The Southern Medical Association

0038-4348/03/9603-0234