Growing reliance on cloud services for storing medical images raises significant privacy concerns. While encryption protects sensitive data, it limits search and retrieval capabilities essential in clinical workflows. Existing methods like additive secret sharing or homomorphic encryption face challenges such as high overhead and limited fault tolerance. We suggest a secret sharing based privacy-preserving image retrieval (PPIR), namely the threshold secret sharing (TSS). Pictures are divided into n shares and distributed across servers, allowing reconstruction with any t or more shares ensuring data availability even during server failures. By integrating TSS with content-based image retrieval (CBIR), our approach enables secure, efficient similarity searches on encrypted medical datasets. Experimental results robust in terms of efficiency, privacy protection and scalability.