Method for storing and managing medical big data by integrating lightweight image classification models

To solve the current problem of large-scale and multi-modal medical data storage and management, this study proposes a medical big data storage and management method that integrates lightweight image classification models. This method innovatively combines lightweight neural networks and attention mechanisms to construct an image classification model, while also building a medical big data storage system and designing corresponding retrieval management schemes. The results showed that the proposed model had accuracies of 0.973 and 0.975, recall rates of 0.95 and 0.953, and mean average precision values of 0.93 and 0.95 on the chest and stomach electronic computed tomography datasets. The write efficiency and query efficiency of the proposed system have been improved by 20.01 and 2.5 times, respectively, with a data compression rate of 53.1%. The hit rate of the proposed solution has increased by 46.7%, while data access and retrieval latency have been reduced by 55.1% and 30.8%. Research has shown that this method significantly improves image classification prediction accuracy, data storage capacity, and data retrieval access efficiency. Research methods can provide storage and management support for multi-modal medical big data, thereby promoting the development of intelligent medical services towards higher quality.