Progress in anatomical tissue imaging, alongside the development of advanced unsupervised learning algorithms for computational analysis, has unlocked substantial potential for enhanced characterization of tissue microstructure. 1 , 2 While data-driven identification of cell clusters reduces subjective bias, the discovery of microanatomical domains often pertains to normal organs with inherent organization. Defining topological principles in tissues without strong structural patterning, such as cancer, has been limited, and deeper stratification in terms of identity and biological function, especially with treatment response, is challenging. Direct evidence and a side-by-side measure of therapy-induced changes could enable differential analysis development for patient stratification and personalized cancer medicine.
Jakubik et al. (Thu,) studied this question.