Early cancer detection has significant clinical value, but there remains no single method that can comprehensively identify multiple types of early-stage cancer. Researchers at EXoPERT Corporation report the diagnostic accuracy of simultaneous detection of 6 types of early-stage cancers (lung, breast, colon, liver, pancreas, and stomach) by analyzing surface-enhanced Raman spectroscopy profiles of exosomes using artificial intelligence in a retrospective study design. It includes classification models that recognize signal patterns of plasma exosomes to identify both their presence and tissues of origin. Using 520 test samples, the system identified cancer presence with an area under the curve value of 0.970. Moreover, the system classified the tumor organ type of 278 early-stage cancer patients with a mean area under the curve of 0.945. The final integrated decision model showed a sensitivity of 90.2% at a specificity of 94.4% while predicting the tumor organ of 72% of positive patients. Since our method utilizes a non-specific analysis of Raman signatures, its diagnostic scope could potentially be expanded to include other diseases.
One test-multi cancer using exosome-SERS-AI
a Overview. Exosome suspension is dropped onto an Au nanoparticle-aggregated array chip and thoroughly dried. Signals were observed at 100 spots (10 × 10) per sample and analyzed by AI algorithms. The system outputs predictions about cancer presence and tissue of origin. A heat map shows actual examples of the representative predicted results for each cancer status. b AI framework. In the first step, diagnostic scores are assigned as the mean values of the multiple instance learning (MIL)-based cancer classifier results. In the second step, signals predicted by the previous cancer classifier are analyzed, then an average score is calculated using six types of prediction models. Cartoons were created with BioRender.com.