This paper presents a volatile organic compound (VOC) gas preconcentration method based on microstructures. The microstructures have a flower-leaf-type pattern, and each microstructure consists of carbon nanotubes (CNTs). This pattern is optimized by performing a simulation study to maximize the adsorption area between the VOC gas flow and the microstructure surfaces. The VOC gases contained in the exhaled breath of a human or dog can be utilized as biomarkers of cancers. Therefore, trace-level VOC gas detection in exhaled breath can enable the early diagnosis of human or canine cancer. We propose a method for analyzing human or canine breath using CNT-based gas preconcentration and conventional gas sensors or gas analysis equipment. To validate the method, we fabricated a CNT-based micro-preconcentrator and experimentally investigated ppb-level gas adsorption and detection with conventional gas chromatography equipment. As a result, we achieved more than 1000-fold gas preconcentration and ppb-level gas detection using conventional gas chromatography, which can normally only detect ppm-level gases.