Introduction to Anticipatory Oncology: From Hidden Disease States to Personalized Medicine

Philip De Melo, Marie St Rose, Mildred Jackson.

Abstract

Cancer develops through continuous and complex biological processes, yet oncology primarily relies on measurements obtained during discrete clinical encounters. Imaging, pathology, biomarkers, molecular tests, and clinical observations provide only partial and sometimes conflicting evidence of the underlying disease process. Although conventional artificial intelligence can predict clinical outcomes from observed variables, it does not necessarily reconstruct the hidden biological state that produces them.

This paper introduces anticipatory oncology as a new paradigm centered on the reconstruction of latent, or hidden, cancer states. It presents latent state artificial intelligence (LAT-AI), as an integrative framework that combines clinical, imaging, pathological, biomarker, and molecular evidence to estimate a clinically interpretable representation of a patient’s underlying disease state. By distinguishing observed measurements from the latent disease process, LAT-AI may:1) Identify ambiguous or discordant evidence, 2) quantify uncertainty, 3) detect clinically meaningful changes, and 4) improve understanding of disease progression.

Anticipatory oncology shifts the central question from “What outcome is associated with the patient’s observed data?” to “What hidden disease state most plausibly generated these observations, and how is that state changing over time?” The framework is intended to complement rather than replace biopsy, pathology, established diagnostic procedures, or clinical judgment. Its goal is to advance precision and personalized medicine by providing interpretable, uncertainty aware, and dynamically updated estimates of hidden cancer states. The paper also examines validation, fairness, data quality, and clinical governance as essential requirements for translating latent state reconstruction into safe and clinically useful decision support.