Full proposal will rely on the following planning:
1. The Oral Medicine units will create a repository of reference images depicting oral lesions: by means of standard photo-camera and smartphone, approximately 500 photos/Unit, will be taken, in accordance with a specific acquisition protocol to provide the best data while limiting the introduction of pictorial artefacts. After the correct conjunction of photo with a diagnosis, all the data collected will be available among the research units in form of online database.
2. The Computer Science research unit will develop innovative algorithms suited for artefact removal in achieved images of the oral mucosa as well as image segmentation, and lesion recognition. Segmentation will be based mainly on colour, texture, and shape analysis on the basis of the nature of the data. The most suitable segmentation methods are described in this table.
After, recognition will be performed using either machine learning approaches or informed techniques such as rule based methodologies. The algorithms developed in the project will recognize the oral mucosal lesions on the basis of the acquired images and the ones already present in the image repertories, if available. Results will be evaluated by the Oral Medicine units.
3. The Oral Medicine units will start testing the experimental prototype in their clinical practice (approximately 50 patients for Unit). The project will release the fully functioning software prototype, and the on-line database as its main outcome along with scientific publications related to the dissemination of both the innovative algorithmic techniques and the medical advances in the diagnosis and care of oral lesions.
Up to date, no existing product is directly comparable to the SEE_DOC (System Expert Ederly Diseases of Oral Cavity). This project represents an innovative evolution of yet existing computer-aided medical prediction system. This system will allow to standardize the diagnostic accuracy of a practitioners/general dentist to that of an oral medicine specialist. The immediate recognition of clinical images of oral mucosal disease, could eliminate all the possible drawbacks related to a second consultation improving the initial diagnostic process, especially in terms of duration. Moreover, the future possibility that the image is captured by the same patient, could dramatically reduce the time related to patient delay, so influential on late diagnosis of OPMD/OSCC.