Transcribed gene fusions are key biomarkers in many hematologic and solid tumors, often representing the primary oncogenic driver mutation. Here, we report an experimental and computational pipeline for detecting fusion transcripts using single-molecule RNA FISH and unbiased correlation analysis (FuseFISH). We constructed a genome-wide database of optimal oligonucleotide sequences, enabling quick design of FuseFISH probes against known and novel fusions. We implemented FuseFISH in cell lines, tissue sections, and purified RNA, reliably detecting one BCR-ABL1 positive in 10,000 negative cells. In 34 hematologic samples, we detected BCR-ABL1 transcripts with high specificity and sensitivity. Finally, we measured BCR-ABL1 expression heterogeneity and dynamics in single CML cells exposed to the kinase inhibitor Nilotinib. Our resource and methods are ideal for streamlined validation of fusions newly identified by next-generation sequencing, and they pave the way to studying the impact of fusion expression variability on clinical outcome.