New evidence suggests a widening toolbox for predicting disease state using imaging and liquid biopsy signals. A meta-analysis of circulating tumor DNA shows post-neoadjuvant ctDNA detectability strongly predicts worse survival across many solid tumors, supporting ctDNA as a high-impact decision variable. Meanwhile, a multicenter study reports that machine learning analysis of routine CT scans can non-invasively predict microsatellite instability status in gastric cancer with clinically useful accuracy, highlighting growing efforts to reduce biopsy burden and accelerate treatment stratification.