Dear Dr. Tim Noeke, First of all, we would like to express our sincere gratitude for your continued trust and use of WEBCEPH over the past three years. Regarding the issue you raised, we would like to inform you that there have been no recent changes made to the AI model used in WEBCEPH for cephalometric analysis. However, we would like to ask if the cephalometric images you upload are consistently captured under the same size and conditions. Upon reviewing the example you provided, it appears that the image may be somewhat close-up, particularly with the occipital region cropped, which is different from the images used during the training of our AI model. WEBCEPH’s landmark detection process follows three steps: 1) Facial region detection 2) General landmark detection 3) Refinement of each landmark within a localized region. If the uploaded image is too close-up, the initial facial region detection may result in a narrower detection range. This can lead to posterior landmarks, such as articulare, being incorrectly positioned more anteriorly than usual. If possible, we recommend using images with a broader region of interest (ROI) during acquisition to ensure more accurate landmark detection. We deeply appreciate your continued use of WEBCEPH and are committed to providing the most accurate and reliable service to support your clinical practice. Should you have any further questions or require assistance, please don’t hesitate to contact us. Sincerely, WEBCEPH Support Team