Temporomandibular disorders (TMD) intelligent diagnosis promises to elevate clinical efficiency, facilitating timely TMD management for patients and staving off the emergence of severe TMJ disability and pain. However, development of TMD intelligent diagnostic tools with high accuracy and sensitivity faces challenges, particularly in sensing minute deformations and ensuring rapid self-healing. Here, inspired by clinical joint palpation, based on the minute facial deformations (0.31-1.62 mm) caused by subtle TMJ movements, we report a biocompatible hydrogel electronic sensor that is capable of instantaneous self-healing (within 2.1 s) and ultra-low detection limit (0.005% strain). The clinical trials involving 120 cases demonstrated that it can efficiently identify the typical clinical manifestations of TMD, such as joint clicking, the adjunctive diagnosis of TMD with repeatability (Kappa=0.86) and a satisfactory accuracy of 90.00%, and it also had a clear indication of other uncoordinated TMJ activity and the timing of TMJ clicking, with a sensitivity of up to 100% for the diagnosis of TMD. Furthermore, to mitigate the reliance on physicians for the diagnostic tool, a predictive model based on sensing waveform features was constructed, achieving high accuracy of 84.4% and sensitivity of 0.86. In summary, the hydrogel sensor, to the best of our knowledge, is the first hydrogel-based strain sensor used for TMD monitoring and is expected to become a radiation-free, non-invasive, convenient and efficient examination tool for clinical TMD diagnosis in the future.