| { |
| "field_id": "F033", |
| "field": "clinical natural language processing", |
| "directions": [ |
| { |
| "direction_id": "G0", |
| "name": "Foundations for clinical language learning", |
| "scope": "Establish domain language representations usable for medical text tasks, or obtain usable training supervision when expert annotations are scarce. Both paths build learning capacity; a single study need not address both representation and annotation." |
| }, |
| { |
| "direction_id": "G1", |
| "name": "Structuring clinical information and terminology", |
| "scope": "Convert clinical or biomedical text into identifiable, usable structured medical information, encompassing facts grounded in the text, standardized concepts for mentions, and standardized codes for documents or patient records. These outputs are not prerequisites for one another." |
| }, |
| { |
| "direction_id": "G2", |
| "name": "Clinical state and functioning assessment from language", |
| "scope": "Use clinical narratives, patient writing, conversations, or speech and language performance to infer patient states, functioning, symptom severity, or future risk. Structured patient data may also be incorporated; the branches need not use the same inputs." |
| }, |
| { |
| "direction_id": "G3", |
| "name": "Clinical NLP capability and reliability assessment", |
| "scope": "Center comparative assessment, external validation, or failure diagnosis to determine whether clinical and biomedical NLP systems are reliable across tasks, prompts, institutions, populations, or clinical judgment conditions." |
| }, |
| { |
| "direction_id": "G4", |
| "name": "Evidence-grounded clinical inference", |
| "scope": "Retrieve, select, or evaluate source evidence and its relationship of support to specific clinical assertions, condition inferences, or code predictions." |
| }, |
| { |
| "direction_id": "G5", |
| "name": "Generation of clinical documentation", |
| "scope": "Turn clinical information from encounter dialogues, existing written records, or medical images into notes, summaries, impressions, or imaging reports for clinical use. These input sources are parallel branches, not jointly required." |
| }, |
| { |
| "direction_id": "G6", |
| "name": "Patient-understandable medical communication", |
| "scope": "Transform medical terminology, reports, or instructional content into explanations, simplified text, or educational interactions that patients can understand and use." |
| }, |
| { |
| "direction_id": "G7", |
| "name": "Privacy-preserving use of clinical text", |
| "scope": "Reduce the risk of patient information disclosure or unequal privacy protection when using clinical text for NLP." |
| }, |
| { |
| "direction_id": "G8", |
| "name": "Text-driven clinical cohort identification", |
| "scope": "Turn trial eligibility criteria and patient information into an operational screening process to identify research cohort candidates or match patients to trials." |
| } |
| ] |
| } |
|
|
