Proposed ultrasound algorithm to differentiate thyroglossal duct and dermoid cysts

1) To assess reproducibility of the previously established SIST score. 2) To determine inter-observer agreement in using ultrasound (US) characteristics to differentiate thyroglossal duct cyst (TGDC) from dermoid cysts (DC) 3) Improve the method used to pre-operatively differentiate TGDC from DC.

Retrospective chart review.

Tertiary care pediatric hospital. SUBJECTS AND METHODS: An electronic medical record was queried to identify children with midline neck masses who underwent pre-operative neck US and had a histopathologic diagnosis of either TGDC or DC. Two pediatric radiologists, blinded to the pathologic diagnosis, evaluated the US images and documented the presence of pre-determined characteristics of each mass. Potential differentiating factors were analyzed for their predictive power. The SIST (septations, irregular walls, solid components = TGDC) score was determined as well as inter-observer agreement. Using the characteristics that had significant predictive power, we used the data to develop an algorithm to improve predicting cyst type.

Pathologically, there were 47 TGDC and 25 DC. The inter-observer agreement about the pathologic diagnosis between the two radiologists was substantial, K = 0.66. Overall, the SIST score predicted the correct diagnosis 67% of the time. Radiologist 1 and radiologist 2 were more accurate than the SIST score alone, making the correct diagnosis 96% and 86% of the time, respectively. In our study, we found that the most important US characteristics in differentiating TGDC and DC are: internal Septations, depth relative to Strap muscles, Shape and Solid parts (4 S algorithm). The SIST score criteria were individually shown to be significant and sensitive in recognizing DC, however, they were not specific and often misclassified TGDC as DC. We developed a new sequential filtering algorithm that more accurately differentiates cysts. This new algorithm uses step-wise filtering of characteristics, first for Septations, then for depth to Straps, then Shape of the cyst and lastly Solid parts (4 S algorithm). This algorithm correctly categorized cyst type in 100% of patients in our study.

Pre-operatively differentiating TGDC and DC continues to be a challenge. Using our 4 S algorithm, we can more definitively differentiate TGDC from DC compared to the SIST score. All SIST score characteristics were significant and sensitive in detecting dermoid cysts, however, not very specific. The radiologists' judgment and accuracy was better than the SIST score. The 4 S algorithm uses sequential filtering of important characteristics: Septations, depth to Straps, Shape of cyst and lastly Solid parts to improve diagnostic accuracy.