Objective

2

To enhance the intelligent operation and maintenance level of magnetic resonance imaging （MRI） equipment， this paper designs and implements an MRI monitoring and early warning system that integrates multi-modal visual recognition algorithms with an edge-cloud collaborative architecture.

Methods

2

The system adopts a three-layer architecture comprising a data acquisition layer， an edge computing layer， and a cloud platform layer. It includes functions such as image acquisition， intelligent recognition， alert notification， parameter configuration， and data management. For image processing， C++ and OpenCV are used to implement dial and color recognition algorithms， while Python and EasyOCR are employed to construct OCR character recognition modules. Python codes are compiled into executable files callable by the main program， ensuring efficient inter-module communication. All algorithms support parameterized configuration to adapt to various device models and display formats. By adopting a “one-time labeling + parameterized configuration” approach， the system reduces computational complexity and improves recognition accuracy. Recognition and alert data are uploaded to the cloud platform， enabling centralized management across multiple hospital campuses. The system is deployed in two campuses of a hospital for trial operation to verify its performance.

Results

2

The solution is low-cost， flexible to deploy， and does not require modification of existing device systems. In terms of monitoring scope， the system can accurately track key parameters such as power status， liquid nitrogen pressure， and environmental temperature and humidity， and provide real-time alerts. Actual test results indicate that power status recognition accuracy reach 100%，liquid nitrogen pressure reading error is within ±0.1，alarm indicator recognition accuracy reach 100%，and environmental temperature display OCR recognition accuracy reach 95%.The system performs image polling every 5 seconds， increasing inspection frequency from once every few days to once per day， thereby significantly improving maintenance efficiency and fault response speed.

Conclusions

2

This system fully integrates multi-modal image recognition with edge–cloud collaboration technology. It features high generalizability， low deployment cost， and strong scalability. The system effectively expands the monitoring scope of MRI equipment， provides a viable solution for remote supervision and centralized management of MRI devices across campuses and brands， and offers reference value for the intelligent maintenance of other high-end medical equipment. It possesses substantial engineering applicability and academic research significance.