The Aim of the study is to evaluate Accuracy of automated mandibular defect reconstruction using Artificial intelligence and assessing impact on aesthetic and occlusion outcomes using patient-specific reconstruction plates.

The digital surgical process often requires an expected mandibular reference model. Currently, the common digital surgery process, is to mirror repair or manually look for other similar mandibles for local data fusion and smoothing processing. A more accurate expected reference model is difficult to achieve, time consuming and difficult to promote in clinical practice. Moreover, rapid routing processing often has poor accuracy. For cumulative bilateral lesions, massive lesions, obvious displacement or lesions cross the middle line, there is still no effective method to predict the expected reference model in clinical practice.

The main objective for conducting this study is to propose an improved algorithm to overcome the drawbacks of recent studies using 3D Unet and to test the predictability and clinical value of virtually generated 3d models of defected mandible in real patients.

Inclusion Criteria:

• Patients with mandibular tumors, cysts or any benign disease resulting in mandibular continuity defect.
• Age group: from 18 - 55 years old.
• No sex predilection.
• CTs or CBCTs of only healthy mandibles from an online database and real data.

Exclusion Criteria:

• Patients with mandibular malignant lesions.
• Children age group from 2-17.
• CTs Of maxilla.
• Elderly patients to be excluded due to the normal physiologic bony change.