Abstract
Over recent years, machine learning and artificial intelligence have become critical components of many cancer detection tests, in particular multi-omic tests such as spectroscopic liquid biopsies. The complexity and multi-variate nature of spectral datasets makes machine learning invaluable in uncovering patterns that enable robust differentiation of cancer signals. However, introducing any AI-enabled medical device into clinical practice is challenging due to the regulatory requirements needed to progress from fundamental research to clinical and patient use. This review explores some of the fundamental concerns in bringing spectroscopic liquid biopsies to the clinic, including the need for explainable artificial intelligence and diverse validation sets. Addressing these issues is essential to accelerate clinical uptake with the ultimate goal of improving patient survival and quality of life.