Automatic measurement of organ volume using ultrasound images

A method is presented for performing the accurate and automatic measurement of the volume of an organ which is full of liquid, such as blood and urine in the cavity. In this method first, using a neural network, each voxel of the 3-D ultrasound data is classified in one of three categories: soft tissue, liquid and boundary. However, low-echo areas of soft tissue tend to be erroneously classified as liquid. Therefore, from those classification results we proposed a new method for measuring the area of the region of interest. We called the new method the "bubble filling method". This method is like the inflation of many balloons in the region of interest. We measured the volumes of agar gel phantoms and fetal bladders to verify the effectiveness of this proposed method. A comparison of the measurements obtained with the manual technique and with the automated technique is included.