Noninvasive quantitative assessment of oral tongue cancer by intraoral ultrasonography

BACKGROUND: To assess tissue characterization of oral tongue cancer and prediction of subclinical cervical lymph node metastasis, we investigated whether intraoral ultrasonography could be used in conjunction with a computer-aided diagnosis (CAD) system. METHODS: The study population comprised 109 patients with presurgical, clinical T1N0 or T2N0 oral tongue squamous cell carcinoma who underwent partial glossectomy. All the patients were examined by preoperative intraoral and postoperative ex vivo ultrasonography. To evaluate the ultrasonic images quantitatively, ultrasonographic parameters from tumor contour features were computed by using the proposed CAD system. The imaging results were correlated with histopathologic findings. RESULTS: Oral tongue cancer was clearly identified in all patients by intraoral ultrasonography. Ultrasonic images of oral tongue cancer reflected the histopathologic structures. Subclinical cervical lymph node metastasis was predicted by intraoral ultrasonography. In a logistic regression analysis using the proposed CAD system, the diagnostic sensitivity, specificity, and accuracy for prediction of subclinical lymph node metastasis were 87.2%, 84.3%, and 85.3%, respectively. CONCLUSIONS: Intraoral ultrasonography in conjunction with the proposed CAD system allows tissue characterization and prediction of subclinical cervical lymph node metastasis.