Potential Applications and Areas of Health-Care Research for ChatGPT and Similar LLMs
| Area no. | Description |
|---|---|
| 1 | Models and applications that can leverage multimodal data such as merging language and imaging, for example, highlighting anomalies in a natural way (with language) when reading PET images |
| 2 | Summary of complex medical histories and records |
| 3 | Summary of information from medical congresses/clinical trial results |
| 4 | Structuring/making information interoperable, for example, during medical documentation (20) |
| 5 | Facilitating clinical documentation such as writing discharge report; once we have structured information, is there really a need for free text? (facts should be communicated reliably and concisely) |
| 6 | Integration with hospital information systems to incorporate patient data, specifications, and requirements (institutional, payer) and resources (staff capacity, provider) |
| 7 | Interpretation and explanation of other AI algorithms (1) |
| 8 | Translation into other languages, with big potential for less frequently used languages for which use of natural language processing was limited in the past |
| 9 | Translation into patient-comprehensible language, making medical information communication more consumer-friendly |
| 10 | Anamnesis |
| 11 | Relief for nursing staff through automized ward communication |
| 12 | Medical writing (21) |
| 13 | Anonymization of clinical text |
| 14 | Fairness, bias in LLMs |
| 15 | Human-in-loop and human-centered design of LLM applications |
| 16 | Chain-of-thought and automated reasoning on LLMs |