Usefulness of Cervical Computed Tomography Hounsfield Units to Differentiate Kawasaki Disease From a Deep-neck Abscess : The Pediatric Infectious Disease Journal

Secondary Logo

Journal Logo

Brief Reports

Usefulness of Cervical Computed Tomography Hounsfield Units to Differentiate Kawasaki Disease From a Deep-neck Abscess

Ogawa, Masato MD, PhD*,†; Matsuki, Yuichi MD, PhD; Hoshina, Takayuki MD, PhD; Kojiro, Masumi MD*; Kusuhara, Koichi MD, PhD

Author Information
The Pediatric Infectious Disease Journal 42(2):p e50-e51, February 2023. | DOI: 10.1097/INF.0000000000003761


We measured the Hounsfield units (HUs) value of cervical plain computed tomography images to differentiate between Kawasaki disease (KD) and a deep-neck abscess (DNA). The HUs value was significantly lower in KD than in DNA, making it a useful marker for differentiating between these 2 diseases.

Kawasaki disease (KD) is a representative febrile illness in childhood. The diagnosis of KD is based on clinical features; therefore, KD cannot be diagnosed until the appearance of at least 5 of the 6 symptoms included in the diagnostic criteria. Some children finally diagnosed with KD present with a only fever and cervical lymphadenopathy and sore throat at the first visit and are often subjected to cervical computed tomography (CT) for the differentiation from a deep-neck abscess (DNA).1–3 The cervical CT image typically shows edematous changes in the posterior wall of the pharynx in KD patients,4 whereas DNA is generally shows rim enhancement (RE) surrounding the abscess on contrast-enhanced CT images.5 However, the accuracy of the detection of RE in the diagnosis of abscess is not very high.6 Furthermore, performing contrast-enhanced CT is often difficult for children with allergic disorders.

Hounsfield units (HU) reflects the radiodensity of the tissue in CT images. This value is defined as 0 in water, –1000 in air and +1000 in bone.7 A previous study reported that the measurement of the HU values in retropharyngeal lesions was useful for differentiating KD from abscess.2 Both contrast-enhanced and plain CT images were used in the previous study, but no report has evaluated the HU value of deep-neck lesions in KD using only plain CT.

We therefore investigated the utility of the HU value measured on cervical plain CT for differentiating between lesions of KD and DNA.


Study Population

A total of 307 KD and 11 DNA patients were hospitalized in the Department of Pediatrics at Kitakyushu General Hospital or the Hospital of the University of Occupational and Environmental Health, Japan, from 2010 to 2020. Among these patients, 17 KD and 8 DNA patients who presented with only a fever and enlarged cervical lymph nodes and a sore throat on admission and who underwent cervical plain CT before their final diagnosis were enrolled in this retrospective study.

The Diagnostic Guidelines for Kawasaki disease (fifth version) were applied as the diagnostic criteria for KD. The enrolled KD patients were clinically suspected of having DNA rather than KD on admission and underwent plain cervical CT, because they had only a fever and cervical lymphadenopathy and/or sore throat. Patients with incomplete KD were not included in this study. Febrile patients accompanied by a sore throat and enlarged cervical lymph nodes were diagnosed with DNA when cervical CT showed low-density areas (LDAs) in the deep-neck region and their symptoms resolved with antimicrobial therapy. None of the DNA patients enrolled in this study were suspected of having KD.

Our study was approved by the Institutional Review Board of Kitakyushu General Hospital and the University of Occupational and Environmental Health, Japan.

Measuring the HU Value

We measured the HU value of LDA showed on plain cervical CT images. All patients were imaged in the supine position over the entire neck using a conventional CT system (Canon Aquilion ONE CT or Canon Aquilion RPIME). The median kilovoltage peak used for CT imaging was 120 (range: 80–120) and the median amperage was 64 mA (range: 20–340). The slice thickness for CT imaging was 1, 3, and 5 mm in 6, 14, and 5 cases, respectively. The median window width to determine LDA of the lesion was 325 (range: 293–595), and the median window level was set at 35 (range: 35–135). Part of the LDA that was sufficiently distant from the normal tissue on CT imaging was enclosed with an ellipse, and the mean HU value in the ellipse was measured (Figure, Supplemental Digital Content 1, The area of the ellipse was unified to 1 cm2. In our hospitals, all CT images are routinely interpreted by multiple radiologists. With reference to the interpretations, in this study, an attending radiologist with over 25 years’ experience determined whether the LDA shown on CT images was suspected of being an abscess or edematous lesion and mainly set the ellipse in the LDA on the cervical CT images. When the radiologist interpreted the CT images, he was blinded to the ultimate diagnosis.

Statistical Analyses

The SPSS statistics software program (version 27; SPSS Inc., Chicago, IL, United States, and IBM, Armonk, NY, United States) was used for all analyses. The Mann-Whitney U test and the Fisher’s exact test were used to compare quantitative and qualitative values, respectively. The cutoff point of the HU value to appropriately differentiate KD from DNA was set using the receiver operating characteristics (ROC) curve. P values of <0.05 were considered to indicate statistical significance.


The demographic and clinical characteristics of patients with KD and DNA are shown in Table 1. There were no significant differences of the age or the incidences of the clinical symptoms between the 2 groups. The serum C-reactive protein levels, but not peripheral white blood cell counts, were significantly higher in KD patients than in DNA patients (P = 0.019). All children received antimicrobial therapy. All KD patients enrolled in this study received antimicrobial therapy after undergoing cervical CT (median, day 3 of illness; range, days 2–5), and the therapy was stopped simultaneously with the diagnosis of KD (median, day 5 of illness; range, days 4–8). Four DNA patients (50%) underwent surgical procedures including the puncture of the lesion and drainage whereas only 1 KD patient underwent the puncture of the lesion. The median day of illness when cervical CT was performed was 3 in both 2 groups.

TABLE 1. - Demographic and Clinical Characteristics of Patients With Kawasaki Disease and Deep-neck Abscess
Patients With Kawasaki Disease, n = 17 Patients With Deep-neck Abscess, n = 8 P
Age, mo (range)*1 52.3 (9.8–160) 57.4 (21.4–84.1) 0.44
Gender, number of male (%) 9 (52.9) 3 (37.5) 0.47
Fever, n (%) 17 (100) 8 (100) 0.16
Sore throat, n (%) 14 (82.3) 6 (75.0) 0.41
Enlarged cervical lymph node, n (%) 17 (100) 7 (87.5) 0.13
White blood cell counts, ×109/L (range)* 23.1 (14.4–34.1) 19.0 (11.9–29.5) 0.199
Serum C-reactive protein level, mg/dL (range)* 11.2 (6.7–27.9) 5.9 (1.1–15.4) 0.019
Antimicrobial therapy, n (%) 17 (100) 8 (100) 1.00
Surgical procedure, n (%) 1 (5.9) 4 (50.0) 0.019
Day of illness when the CT scan was taken, days (range)* 3 (2–5) 3 (1–7) 0.84
Hounsfield units value (range)* 16.4 (7–27) 28.5 (19.8–47) 0.0046
*Median value (interquartile range).

Cervical CT HU values in KD patients (median: 16.4, range: 7–27) were significantly lower than those in DNA patients (median: 28.5, range: 19.8–47.0) (P = 0.0046) (Table 1, Figure, Supplemental Digital Content 2, To evaluate the accuracy of the HU value for differentiating KD from DNA, we created an ROC curve (Figure, Supplemental Digital Content 3, The area under the ROC curve for the HU value was 0.93 (95% confidence interval, 0.83–1.03). The HU value of <28.0 had a sensitivity of 75.0% and specificity of 100%.


This study indicated that, in deep-neck lesions detected by cervical plain CT, the HU value might be useful for differentiating KD from DNA. In KD patients, the incidence of LDAs showed on cervical CT is controversial. A previous report found that deep-neck-space-infection-like symptoms were present in 5% of KD patients with head and neck manifestations.8 However, the incidence of such lesions may be underestimated, as not all KD patients routinely undergo cervical CT,2,3 and pediatricians often experience KD patients with LDAs on CT. Thus, a useful marker for differentiating deep-neck lesions caused by these 2 diseases is needed.

Our study had 2 unique points. First, all KD patients underwent CT before the diagnosis of KD. Some KD patients, especially older children, present with only a fever and cervical lymphadenopathy at admission.1,3 Most of these patients were initially treated with antimicrobial agents, and surgery was occasionally performed for them.3 The early differentiation of KD from DNA by the cervical CT HU value can prevent unnecessary antimicrobial therapy and surgical procedures. Second, unlike previous study using both contrast-enhanced and plain CT,2 only cervical plain CT was used in this study. The present findings suggested that cervical CT for distinguishing deep-neck lesions could be performed more safely without the use of an iodine contrast agent, which can cause anaphylaxis.

The pathology of deep-neck lesions in KD remains unclear. Our findings indicated that deep-neck lesions were likely to be found from the early acute phase of KD, as cervical CT was performed by day 5 of illness in all KD patients. Given the present results and the fact that KD is a systemic vasculitis, the deep-neck lesion may reflect perivascular edema and the dilation of venous capillaries. LDAs shown in abscesses are considered to indicate the infiltration of large amounts of leukocytes to soft tissues.9 As the HU value reflects the radiodensity of the tissue on CT, and given that this value is defined as 0 in water,7 the HU value may be lower in the KD lesions than in DNA ones, as the former are thought to contain fewer leukocytes than the latter.

This study was associated with some limitations. First, the diagnosis of DNA was made based on clinical symptoms and the response to antimicrobial therapy. However, we believe that the diagnosis of DNA was correct because the disappearance of the LDA in the recovery phase was confirmed in most patients. Second, in this study, only 1 radiologist interpreted the CT images. The personal habits and beliefs of this radiologist may have affected the research results. Finally, the study population was relatively small, which could have affected the accuracy of the statistical analysis. In addition, a small study may be considered insufficient to justify the conclusion of the study. To resolve the second and third limitations, a large-scale study involving multiple radiologists interpreting the CT images is desired to confirm our results.

In conclusion, the measurement of the cervical plain CT HU value in deep-neck lesions can be useful for differentiating KD from DNA. Such differentiation can allow appropriate therapeutic interventions to be performed for KD patients, avoiding unnecessary antimicrobial therapy and surgical procedures.


We appreciate the help of Dr. Brian Quinn (Japan Medical Communication, Fukuoka, Japan) for editing the aricle.


1. Nomura Y, Arata M, Koriyama C, et al. A severe form of Kawasaki disease presenting with only fever and cervical lymphadenopathy at admission. J Pediatr. 2010;156:786–791.
2. Sasaki T, Miyata R, Hatai Y, et al. Hounsfield unit values of retropharyngeal abscess-like lesions seen in Kawasaki disease. Acta Otolaryngol. 2014;134:437–440.
3. Ueda Y, Saita Y, Matsuzawa T, et al. Six patients with Kawasaki disease showing retropharyngeal low-density areas on computed tomography. Pediatr Int. 2010;52:e187–e189.
4. Katsumata N, Aoki J, Tashiro M, et al. Characteristics of cervical computed tomography findings in Kawasaki disease: a single-center experience. J Comput Assist Tomogr. 2013;37:681–685.
5. Boucher C, Dorion D, Fisch C. Retropharyngeal abscesses: a clinical and radiologic correlation. J Otolaryngol. 1999;28:134–137.
6. Vural C, Gungor A, Comerci S. Accuracy of computerized tomography in deep neck infections in the pediatric population. Am J Otolaryngol. 2003;24:143–148.
7. Sudhyadhom A. On the molecular relationship between Hounsfield Unit (HU), mass density, and electron density in computed tomography (CT). PLoS One. 2020;15:e0244861.
8. Yoskovitch A, Tewfik TL, Duffy CM, et al. Head and neck manifestations of Kawasaki disease. Int J Pediatr Otorhinolaryngol. 2000;52:123–129.
9. Mohammad M S, Na M, Hu Z, et al. Staphylococcus aureus lipoproteins promote abscess formation in mice, shielding bacteria from immune killing. Commun Biol. 2021;4:432.

Kawasaki disease; deep-neck abscess; Hounsfield units; plain cervical computed tomography; low density area

Supplemental Digital Content

Copyright © 2022 Wolters Kluwer Health, Inc. All rights reserved.