Thyroid Texture Classification and Volume Monitoring using Autoregressive Modelling and Machine Learning Approaches

Thyroid diseases often involve change in the shape and size of thyroid over time. Hence, we propose a novel approach of thyroid texture classification and finally volume monitoring using Autoregressive (AR) Modelling. Several features are extracted from thyroid and non-thyroid regions using AR modelling to train various machine learning based classifiers which later are used for thyroid texture classification and ultimately segmentation. The segmented 2D thyroid images from a freehand thyroid Ultrasound scan are used in the final stage to produce a 3D thyroid after 3D reconstruction. The volume of this 3D thyroid is monitored over time to diagnose any possible thyroid related diseases.