The flexible sensor is an effective solution to solve the failure of conventional sensors at cryogenic conditions. The electrical capacitance tomography (ECT) phase distribution imaging system for cryogenic two-phase flow was focused on. The sampling stability of flexible sensors based on flexible circuit boards was compared with that of traditional patch sensors. A deep neural network (DNN-LNG) was trained to measure the phase distribution of LNG- saturated methane vapor two phase flow. Numerical simulation experiments were conducted to evaluate the model by comparing it with traditional algorithms. Alternative working medium experiments were conducted to compare the two sensor schemes, and the flexible sensor with a better imaging effect was selected to combine with DNN-LNG. A room-temperature experiment was conducted to substitute polypropylene-air as LNG-saturated methane vapor two-phase flow, testing the performance of the flexible sensor and DNN-LNG algorithm in imaging complex flow patterns. And then the liquid nitrogen imaging test was conducted. Results show that flexible sensors can effectively improve the sampling stability and inversion accuracy of ECT system phase distribution, and its standardized design and manufacturing process as well as the stable high installation accuracy provide a solution for the large-scale manufacturing of ECT sensors. The DNN-LNG imaging network can greatly eliminate artifacts and distortions, and obtain clear and accurate imaging results. The combination of flexible sensors and DNN-LNG imaging method has broad application prospects in LNG two-phase flow imaging.