• Capsules
• Gastrointestinal
• Hand-Held
• localization method
• Real-time
• Small Intestine

Wireless capsule endoscopy is a prevalent method to visualize and diagnose the human gas-trointestinal tract but estimating the exact position of the capsule during the acquisition of each image is a very critical issue and has been a challenge in recent years. Although there are many applications of localization methods in the literature, there are no hand-held device to localize endoscopic capsules. Compared to devices used to get information about the degree of Freedom (DoF) of the capsule, such as robotic arms, the aim of the thesis has been to develop, for the first time, an hand-held system for real-time identification of the 3D configuration of magnetic devices in gastrointestinal procedures.The developed prototype consists of five Hall sensors positioned on the lateral and upper faces of a cube and an Imu sensor positioned in the lower part, there are also a led sensor and a vibrating motor useful to give a visual and vibrotactile feedback respectively. The heart of the electronics is the Argon microcontroller that, exploiting the sending through Wifi, sends magnetic field values in real time to Matlab. The results obtained are encouraging and concern the update rate of the sensors which is of the order of the millisecond, the error of the magnetic field detected by the sensors comparedto the dipole model which is negligible, a low power consumption of the overall system and a recharge time of only three hours using the power of the Wifi charger. The final test with the prototype was to test it with a localization algorithm obtaining results in terms of error of position in the range between 1.3cm and 2.6cm depending on the distance from the magnet (in particular when the magnet was very close, 8cm , the dipole model fails so the error is greater as well as for very large distances, 16cm, where the noise prevails). For intermediate magnet positions such as 10-12 cm the algorithm can detect the position of the magnet with with a smaller error. The trend of the orientation error follows the position error, in a range of 2.5- 3.4 degrees.