Circulating genetically abnormal cells (CACs) serve as crucial biomarkers for the early detection of lung cancer. The identification of CACs in patients holds significant clinical value for the timely diagnosis of lung cancer. Prior research has focused on developing deep learning-based cell segmentation and signal spot detection algorithms to facilitate the detection of patients’ CACs. However, these methods often involve multi-stage processing pipelines, which can be time-consuming. To address this issue, we introduce TRP-Net, a transformer-based end-to-end network designed for the efficient detection of CACs in multi-channel fluorescence in situ hybridization (FISH) images. Experimental results demonstrate that our proposed method achieves performance comparable to the state-of-the-art (SOTA) approach while significantly enhancing the prediction speed by a factor of 5.853.