A hybrid localization method for Wireless Capsule Endoscopy (WCE)

Wireless capsule endoscopy (WCE) is a well-established diagnostic tool for visualizing the Gastrointestinal (GI) tract. WCE provides a unique view of the GI system with minimum discomfort for patients. Doctors can determine the type and severity of abnormality by analyzing the taken images. Early diagnosis helps them act and treat the disease in its earlier stages. However, the location information is missing in the frames. Pictures labeled by their location assist doctors in prescribing suitable medicines. The disease progress can be monitored, and the best treatment can be advised for the patients. Furthermore, at the time of surgery, it indicates the correct position for operation. Several attempts have been performed to localize the WCE accurately. Methods such as Radio frequency (RF), magnetic, image processing, ultrasound, and radiative imaging techniques have been investigated. Each one has its strengths and weaknesses. RF-based and magnetic-based localization methods need an external reference point, such as a belt or box around the patient, which limits their activities and causes discomfort. Computing the location solely based on an external reference could not distinguish between GI motion from capsule motion. Hence, this relative motion causes errors in position estimation. The GI system can move inside the body, while the capsule is stationary inside the intestine. This proposal presents two pose fusion methods, Method 1 and Method 2, that compensate for the relative motion of the GI tract with respect to the body. Method 1 is based on the data fusion from the Inertial measurement unit (IMU) sensor and side wall cameras. The IMU sensor consists of 9 Degree-Of-Freedom (DOF), including a gyroscope, an accelerometer, and a magnetometer to monitor the capsule’s orientation and its heading vector (the heading vector is a three-dimensional vector pointing to the direction of the capsule's head). Four monochromic cameras are placed at the side of the capsule to measure the displacement. The proposed method computes the heading vector using IMU data. Then, the heading vector is fused with displacements to estimate the 3D trajectory. This method has high accuracy in the short term. Meanwhile, due to the accumulation of errors from side wall cameras, the estimated trajectory tends to drift over time. Method 2 was developed to resolve the drifting issue while keeping the same positioning error. The capsule is equipped with four side wall cameras and a magnet. Magnetic localization acquires the capsule’s global position using 9 three-axis Hall effect sensors. However, magnetic localization alone cannot distinguish between the capsule’s and GI tract’s motions. To overcome this issue and increase tracking accuracy, side wall cameras are utilized, which are responsible for measuring the capsule’s movement, not the involuntary motion of the GI system. A complete setup is designed to test the capsule and perform the experiments. The results show that Method 2 has an average position error of only 3.5 mm and can compensate for the GI tract’s relative movements. Furthermore, environmental parameters such as magnetic interference and the nonhomogeneous structure of the GI tract have little influence on our system compared to the available magnetic localization methods. The experiment showed that Method 2 is suitable for localizing the WCE inside the body