This work aimed to develop a fibre optic sensor capable of monitoring the joint movements of the fingers of the hand. This was achieved using multicore optical fibres (MCFs) with fibre Bragg gratings (FBGs) multiplexed along the length of the fibre and in the wavelength. Through the measurement of their spectral wavelength shift, it was possible to reconstruct the directional curvature in multiple points along the length of the MCF. The sensor was characterised in two configurations: using only the MCF and using the MCF encapsulated in a capillary tube. Various materials were tested for the tube, and it was concluded that the most suitable was the nitinol tube due to its superelastic properties, which allow the fibre to return to its original shape after deformation. The sensitivity curves obtained by the radial core-FBGs for different angular positions of the MCF revealed satisfactory results and were similar between each other, demonstrating good performance for the characterisation of the sensor. After obtaining the maximum sensitivities for each FBG inscribed in each MCF core, it was possible to reconstruct the curvature vector at a specific MCF length. Tests of the sensor's performance were carried out using the data obtained for calibrating the sensor, showing that it was capable of reconstructing both the amplitude and direction of the curvature vector with average errors of 5% and 1%, respectively. To validate the robustness of the method, random curvatures and orientations were tested for the four quadrants, obtaining errors lower than those mentioned above. In a second phase of the work, the sensor was used to reconstruct the twodimensional shape of the MCF. It was observed that the reconstructed shape was close to the predicted one, despite localised errors. These were attributed to the distance between the FBGs (≈ 2.5 cm) and the model used for reconstruction which uses the spatial position of the prior FBG to estimate the spatial position of the next, leading consequently to the propagation of errors. In addition, the reconstruction method assumes that the MCF maintains its angular alignment along the MCF length. However, since the MCF length was ≈ 70 cm, small twists can appear along its length, contributing to additional errors. In the final phase of this thesis, the sensor's performance was evaluated in the index finger, using three scenarios: using a 3D printed finger, through a robotic hand, and in the finger of a healthy individual. In each case, the amplitude of the curvature vector was reconstructed and later compared with the predicted one, showing better performance for smoother curvatures. The average errors obtained were in the order of 6% in the case of the tests carried out on the 3D printed finger and 4% in the case of both the healthy individual and the robotic hand. These results showed promising results regarding the use of this technology for the monitoring of the degrees of motor function disability of the hand in rehabilitation contexts, guaranteeing a powerful tool for the occupational therapist.