Automatisation of sono-morphologic staging of cervical lymphnodes in head and neck cancer, a possible alternative?

Objectives: B-scan ultrasound imaging is an obligatory examination of great clinical relevance in patients with head and neck cancer. The procedure, h owever; is dependent on an experienced examiner and thus requires higher pe rsonal ressources than other examination techniques. In addition, its subje ctive character leads to possible deficits in reliability and objectivity. Established sonomorphologic criteria (echo-contour, echo-density, echo-stru cture, boundary) should therefore be matched to the findings of computer ai ded sonomorphometry. Method: 200 sonographic images of the cervical region of patients with oral carcinoma were analysed semi- and fully automatically as well as interactively by standard image analysis software (Quan-timed 5 00, Leica, Germany). The results were evaluated statistically and correlate d to the findings of both an experienced and inexperienced examiner. Result s: Fully automatic image analysis was impossible, as the software did not c orrectly identify the lymph nodes without interactive support. Quality of t he semiautomatic and interactive image analysis mainly depended on the corr ect manual identification of the boundaries of the lymph node. Regarding th is criterion the image analysis software was not able to compensate for the deficits of an inexperienced examiner. Conclusion: Delegation of the sonog raphic analysis to inexperienced, less qualified personnel, even with the h elp of image analysis software, seems to be inadvisable. This is especially the case if three-dimensional information from the real time analysis is n ot available.