X-rays are widely used in diagnostic medical imaging – in this Bachelor thesis, they will be used for automatic age determination. Thereby radiographs of the jaw can provide important clues about the age of the person because dental growth is less influenced by diet and hormones than skeletal growth. Compared with histological and biochemical methods is X-ray imaging significantly faster and facile.
As dental tissue is usually very well preserved after death and remains fairly unchanged for thousands of years, its analysis is widely used in forensics. Age determination on living persons is carried out to determine whether the child has reached the age of criminal responsibility or the majority if the birth certificate is not available.
However, the accuracy of the age determination by physicians is always doubted. On aver- age, the age estimation for children and adolescents differs by about half a year and about two years in the case of particularly serious inaccurate estimates. For adults, the result is usually even less accurate. Therefore, in the context of this bachelor thesis, an attempt will be made to develop a deep learning algorithm for age estimation. Since promising results have already been achieved with Deep Learning in other areas of medical image analysis – automated solutions could support physicians in estimating the age of the patient, in order to achieve more reliable results. The neuronal networks will be trained with a data set of 12 000 panoramic dental X-rays labeled with the age of the patients in days and provided by the University Hospital Erlangen. So the aim is to develop a supervised approach. Since convolutional neural networks (CNNs) have already achieved good results in other areas of medical image analysis [4], they will also be used for this task.