Automatic estimation of age at death from human bone cross-sections using neural networks
T.J. Morgan, Zhigang Liu, Marimuthu Swami Palaniswami · 2002
This paper describes on going research on digitally automating the estimation of age at death of human corpses, using micro-radiograph images of mid-shaft bone cross-sections. Traditional methods used in previous studies have involved manually calculating various statistical measures from a bone cross section. These methods were both time consuming and inaccurate, consequently the databases obtained were usually very small and hence statistical conclusions of a qualitative or quantitative nature were difficult to establish. The proposed automated system is able to process a complete bone cross-section. Various bone micro-structures are segmented using traditional image processing techniques. Statistical measures are extracted from the segmented images. These measures form the feature vectors for classification. Initial results based on two neural network classification algorithms are provided for the estimation of age at death.