AI and Traditional Chinese Medicine: Documentation and Research at WUC

Connect two disciplines around a defined problem

WUC offers MSAHM and MSCSE programs in San Jose, providing an educational setting in which clinical and computing perspectives can meet. The university has confirmed AcuScribe, a classical-text knowledge graph and tongue-image AI as active projects. Their research aims are described here; project-specific evaluations and current participation arrangements should be discussed with WUC.

A useful project begins with a specific task. Drafting a note, finding a passage in a reference work and measuring features in an image are different problems. Each requires its own data, evaluation and safeguards. Combining them under a broad “AI in medicine” label does not establish that a system can diagnose or prescribe.

AcuScribe: a public Beta documentation demonstration

AcuScribe’s public project site presents an ambient speech-to-text scribe and a structured visit-record / PDF exporter. The demonstrated workflow is to transcribe a consultation, review fields such as complaint, tongue and pulse, pattern and formula, correct the generated draft, and export a bilingual English–Chinese PDF. The sample consultation on the project site is fictional.

The project disclaimer identifies it as a Beta technical demonstration, not a certified medical device, electronic health record system or clinical decision-support tool, and not intended for production clinical use. It also states that audio and text are processed by third-party AI services without a healthcare Business Associate Agreement in place. Review those terms before interacting with the demonstration; use appropriate fictional or non-sensitive examples rather than treating it as a clinical record system.

Research questions include whether a draft preserves the meaning of an encounter, omits information or introduces unsupported statements. External ambient-scribe research provides context but does not validate AcuScribe. Accuracy, time savings and suitability for a specific clinical workflow require project-specific evaluation.

Classical-text knowledge graph: retrieve sources, preserve context

The classical-text project explores organizing and retrieving references in Chinese medical literature. A useful retrieval system should expose the exact source passage and distinguish a source quotation from a generated explanation. Different editions, translations and historical terminology can affect interpretation.

Potential evaluation questions include whether a cited passage exists, whether it answers the query and whether the system communicates uncertainty. A retrieval-augmented system can still return irrelevant material or generate an unsupported answer. Historical references should not be converted automatically into contemporary treatment instructions or presented as verified herb–drug interaction advice.

Tongue-image AI: study reproducible image measurements

The tongue-image direction investigates image analysis, where lighting, camera settings, image quality and data labeling can affect results. Research may examine reproducible visual features, but measuring an image feature and establishing a clinically meaningful diagnosis are separate tasks.

Before making performance claims, a project needs a clearly defined task, an appropriate reference standard and evaluation on suitable data beyond the development set. Repeated-image consistency is not equivalent to evidence of treatment benefit. This overview does not claim that a smartphone image determines a patient’s condition or that these models are approved for independent diagnosis.

Evaluate safety, privacy and human review

AI research involving healthcare information needs an appropriate data-governance process. Access to an educational clinic does not give every engineering student unrestricted access to patient records. Project permissions, consent where applicable, access controls, retention and research approvals must be established for the particular activity.

NIST’s AI Risk Management Framework offers a general reference for evaluating trustworthy systems. Practical research questions include how errors are detected, who can correct an output, how performance changes across languages and contexts, and how the system behaves when information is incomplete. A human review step should be meaningful rather than a label added to an otherwise untested workflow.

Ask about project status and educational participation

Prospective students can ask WUC about current project leads, supervision, participation requirements and the evidence available for each initiative. An active project may change scope or availability. Participation is not guaranteed by admission to either degree, and a research overview is not a promised specialization or clinical service.

For MSAHM applicants, the core decision remains the published academic and supervised clinical program. For MSCSE applicants, review the actual curriculum and capstone arrangements. Interdisciplinary interests can guide a faculty conversation, while employment, research outcomes and commercial deployment require separate evidence.

Common questions