Rust is gaining attention as a language that is more memory-safe compared to C, and there is a strong motivation to convert existing C programs to Rust. While previous studies have explored using LLMs for code translation, they have indicated that it is challenging to generate Rust code that successfully compiles, even for programs with a few hundred lines. We proposed a LLM-based translation scheme leveraging parsing data. We observed so far that parsing the original C code and strictly converting it from the referenced elements, and summarizing and including the information of referenced elements (such as called function signatures and definitions of used data types) in the prompt during both the conversion and compile repair phases, improve the compilation success rate and equivalence of the generated code.