Systems for detecting and measuring certain dental defects and/or existing restorations, such as lesions which may or may not invade interproximal areas, using a neural network is disclosed. In some embodiments, a patient's health conditions and certain dental conditions such as crowding, presence of implants, gum disease, etc. are also considered by an automated system to classify the dental health of a patient into one or more risk classes. The system then suggests appropriate remedial measures and an appropriate maintenance program for that patient to keep therapy within guidelines of proven standard of care, leading the way to reduce tooth loss in population.