All research

Clinical AI / Published · 2026

Sequential Pattern Transformer

From clinical history to possible futures

A generative and interpretable framework for predicting disease trajectories.

01 / Inside the method

How it works.

Select a stage to explore the workflow.

01Disease sequences02Mine patterns03Transformer decoder04Next disease codes

Conceptual schematic · Simplified method overview

Clinical histories are represented as ordered sequences of disease codes. Order provides context beyond the presence of individual diagnoses.

PrefixSpan extracts recurring sequential patterns to prepare the input used by the framework.

A decoder learns relationships within the sequence and predicts what could follow the observed history.

The model produces candidate next codes. These predictions describe learned associations, not a certain outcome for an individual.

02 / Context & sources

The research.

Modeling health as a sequence

The Sequential Pattern Transformer (SPT) studies sequences of clinical events to predict disease trajectories. The framework combines generative prediction with interpretable sequential patterns, connecting a patient’s recorded history to possible future diagnoses.

Publication

Mohammad Assadi Shalmani, Masoud Khani, Amirsajjad Taleban, Zihao Yi, Jennifer T. Fink, Christopher E. Weber, Qiang Lu, and Jake Luo. Sequential pattern transformer (SPT): a generative and interpretable framework for predicting disease trajectories. Neural Computing and Applications, 38, 28 (2026).

Read the paper and methods

Explore the source code