Skip to main content
NCBI home page
Mayo Clinic Proceedings: Digital Health logoLink to Mayo Clinic Proceedings: Digital Health
. 2023 Jun 30;1(3):299–308. doi: 10.1016/j.mcpdig.2023.06.002

Plastic Surgery and Artificial Intelligence: How ChatGPT Improved Operation Note Accuracy, Time, and Education

Ahmad M Abdelhady 1,∗, Christopher R Davis 1
PMCID: PMC11975639  PMID: 40206608

Abstract

Objectives

To evaluate the feasibility and effectiveness of employing a large language model (LLM), particularly the chat generative pretrained transformer (ChatGPT), as a supporting and educational tool for plastic surgeons.

Patients and Methods

The study involved generating operative notes for plastic surgery procedures using ChatGPT-4 and comparing them with handwritten or computer-written notes for the same procedures. All operations were performed in a single institution from February 1, 2023, to April 20, 2023, by 4 surgeons. Data were compared using the Likert scale that included the following: procedure type; seniority of operating surgeon and operative note creator; type of note; time of surgeon generated note; time of artificial intelligence (AI)-generated note; adherence of AI note to current guidelines; surgeon satisfaction about AI-generated note; patient demographic characteristics; and patient satisfaction about AI-generated note.

Results

ChatGPT-generated operative notes (n=30) took considerably less time to create than human-generated notes (5.1 seconds vs 7.10 minutes; P<.05), with 100% of the ChatGPT notes adhering to the current guidelines. Surgeons and patients expressed high satisfaction with the process of generating operative notes by the AI with (n=13) surgeons being very satisfied, (n=12) surgeons being satisfied, (n=4) surgeons being neutral, and only (n=1) surgeon was dissatisfied with the generated operative note. From a patient point of view, (n=13) patients were very satisfied, (n=13) patients were satisfied, (n=3) patients were neutral, and only (n=1) patient was dissatisfied. The mean time for surgeons to create a human note was 7.10 minutes (standard error of mean of 0.334 minutes). By contrast, the ChatGPT platform generated notes in 5.1 seconds (standard error of mean of 0.12 seconds), with an average of 2.1 edits needed per note. In addition, another AI tool enhanced the operative notes with realistic image creations to illustrate surgical incisions.

Conclusion

LLM-generated notes are significantly faster to create, adhere to guidelines, and receive high user satisfaction from surgeons and patient perception. AI-generated hyper-realistic images further enhanced the operative notes. Although not able to replace human input, these findings show AI’s potential in the medical field and opportunities for future advancement.

In surgical practice, operative notes are crucial for medical documentation, information exchange, communication among health care professionals, and as a legally binding document. Mistakes or oversights in these notes can have severe implications for patient care and result in misunderstandings. Failure to comprehensively record an operation can result in incomplete medical notes, additional questions from future surgical procedures, and potential medicolegal repercussions. To unify the guidelines for writing a comprehensive operation note, the American College of Surgeons and Royal College of Surgeons issued a list of essential points to include in all surgical procedure operation notes.1

An extensive neural network with billions of weights or more, known as a large language model (LLM), is an artificial intelligence (AI) language model. It uses self-supervised learning to train on large quantities of unlabeled text. LLMs have gained prominence since 2018 and exhibited exceptional performance on multiple tasks, leading to a shift in natural language processing research away from specialized supervised models for tasks. In November 2022, OpenAI software (OpenAI) released chat generative pretrained transformer (ChatGPT),2 an AI-powered LLM that uses the GPT-3 deep learning language models and is continuously refined by supervised and reinforced learning techniques. It was followed by GPT-4 shortly after in March 2023. Unlike conventional search engines that deliver generalized information, ChatGPT has gained substantial attention for its ability to produce personalized answers to specific questions.

Methods

Between February 2023 and April 2023, we conducted a prospective study that compared the speed and accuracy of operative notes generated by a LLM (ChatGPT) with those that were handwritten or typed using a normal text editor on a computer. The notes generated using ChatGPT-4 were completed by creating an account on the current website for the service https://chat.openai.com, and using the direct written prompt at the text field on the ChatGPT website. The direct text prompts used included “Write me an operation note for excision of skin cancer BCC and direct closure with 5/0 Vicryl Rapide sutures” and “Write me an operation note for nail bed repair, done under general anesthesia in a 2-year-old patient. Closure with 6/0 Vicryl Rapide sutures” (Video).

We have also generated AI hyper-realistic images that could be used as a visual aid to be attached to the notes by leveraging another open AI product called DALL-E by signing up to the service using https://openai.com/product/dall-e-2 and entering text prompts of the desired photo into the text field.

The study was conducted on 30 plastic surgery procedures performed at the same institution by 4 surgeons. Procedures were classified as major (n=1) for operations done under general anesthetic or axillary block and performed mainly by a consultant; intermediate (n=11) for operations done under local anesthetic by a consultant or with by a trainee surgeon a consultant supervising in the theater; and minor (n=18) for operations done independently by juniors with consultants as senior surgeons performing (n=5) procedures and trainee junior surgeons performing (n=25) both supervised and unsupervised.

Operative notes were written mostly by junior surgeons who were assisting or operating in these procedures by computer text software using a premade template (n=28) or using traditional pen and paper (n=2).

The following metrics or data points were gathered: procedure name—procedure type—seniority of operating surgeon and operative note creator—type of note—time of surgeon generated note—time of AI-generated notes—adherence of AI note to the current guidelines—surgeon satisfaction about AI-generated note—patient demographic characteristics—patient satisfaction about AI-generated note.

Results

The results of the testing are presented in (Table). The average time taken by human surgeons to create a note was 7.10 minutes with the standard error of mean (SEM) of the data being ∼0.334 minutes. By contrast, the ChatGPT platform generated notes in an average of 5.1 seconds, with SEM of the data being ∼0.12 seconds and an average of 2.1 edits were needed per note, and 100% of the notes generated by the ChatGPT adhered to the guidelines (Figure 1). This indicated that the ChatGPT platform is considerably faster in generating operation notes when compared with humans, with greater accuracy.

Table.

Table Showing Comparison Between Human-Generated Notes and AI-Generated Notes, such as: Procedure Name for example BCC (Basal cell carcinoma) excision—Procedure Type—Seniority of Operating Surgeon and Operative Note Creator—Type of Note—Time of Surgeon Generated note—Time of AI-generated note—Adherence of AI note to the current guidelines—Surgeon satisfaction about AI-generated note—Patient demographic characteristics—Patient satisfaction about AI-generated note

Procedure Procedure Type Operating Surgeon Seniority Seniority of Operation Note Creator Type of Note: Handwritten vs Computer written Computer Notes Template Used Time of Generation by Hand or Computer Time of Generation by Chat GPT Number of Edits Needed to Complete the Operation Note Adherence to the Current Guidelines Evaluation of Quality by the Operating Plastic Surgeon on Likert Scale Patient Age (y) and Sex Patient’s Satisfaction with Operation Notes to be Mainly Generated by AI
1 Excision of BCC on the nasal ala and local flap. Intermediate Senior Senior Hand N/A 12 min 7 sec 2 Edits Yes Very satisfied 45—F Very satisfied
2 Excision of multiple skin lesion and reconstruction with a split thickness skin graft. Intermediate Junior Junior Computer Yes 23 min 6.01
sec		 2 Edits Yes Very satisfied 41—F Satisfied
3 Left breast mastopexy with wise pattern. Major Senior Senior Hand N/A 15 min 4.9 sec 3 Edits Yes Very satisfied 48—F Satisfied
4 Excision of a melanoma scar on the right cheek closed with absorbable sutures. Intermediate Senior Junior Computer Yes 7 min 4.16 sec 3 Edits Yes Very satisfied 28—F Satisfied
5 Closure of wound, right index finger dorsum. Minor Junior Junior Computer Yes 6 min 4.97 sec 4 Edits Yes Very satisfied 46—M Very satisfied
6 Extraction of metal foreign body right index finger. Intermediate Junior Junior Computer Yes 8 min 5.4 sec 2 Edits Yes Very satisfied 44—F Satisfied
7 Repair of fingertip laceration. Minor Junior Junior Computer Yes 11 min 4.5 sec 2 Edits Yes Very satisfied 31—M Very satisfied
8 Terminalisation of right thumb osteomyelitis. Intermediate Junior Junior Computer Yes 9 min 13.9 sec 4 Edits Yes Very satisfied 38—M Satisfied
9 Repair of a left-hand laceration and ulnar digital nerve repair. Intermediate Junior Junior Computer Yes 8 min 6.08 sec 3 Edits Yes Very satisfied 24—M Very satisfied
10 Repair of a laceration to the dorsum of the left hand and repair of superficial branch of radial nerve. Intermediate Junior Junior Computer Yes 6 min 4.05 sec 2 Edits Yes Very satisfied 22—M Very satisfied
11 Washout and foreign body extraction of the right index finger. Minor Junior Junior Computer Yes 5 min 4.83 sec 3 Edits Yes Very satisfied 48—F Very satisfied
12 Nailbed repair and closure of pulp laceration left index finger. Minor Junior Junior Computer Yes 7 min 5.35 sec 1 Edit Yes Very Satisfied 33—F Very Satisfied
13 Nailbed repair and closure of pulp laceration left index finger. Minor Junior Junior Computer Yes 4 min 5.30 sec 1 Edit Yes Very Satisfied 30—F Very Satisfied
14 Right middle finger dorsal wound laceration closure. Minor Junior Junior Computer Yes 5 min 4.42 sec 4 Edits Yes Neither satisfied nor dissatisfied 51–M Neither satisfied nor dissatisfied
15 Extensor tendon repair right ring finger. Minor Junior Junior Computer Yes 8 min 4.13 sec 3 Edits Yes Satisfied 41—F Satisfied
16 Excision of lesion on left lower leg and direct closure. Minor Junior Junior Computer Yes 8 min 4.8 sec 3 Edits Yes Satisfied 85—F Neither satisfied nor dissatisfied
17 Excision of lesion on dorsum right hand. Minor Junior Junior Computer Yes 6 min 4.71 sec 2 Edits Yes Satisfied 56—M Very satisfied
18 Punch biopsy of lesion 3mm nasal tip. Minor Junior Junior Computer Yes 6 min 4.58 sec 1 Edit Yes Satisfied 51—F Neither satisfied nor dissatisfied
19 Excision of giant cell tumor on left palm. Intermediate Senior Junior Computer Yes 11 min 4.65 sec 2 Edits Yes Dissatisfied 38—M Satisfied
20 Left carpal tunnel decompression. Intermediate Senior Junior Computer Yes 12 min 4.73 sec 3 Edits Yes Neither satisfied nor dissatisfied 44—F Satisfied
21 Excision of lesion left on inferior side of the breast. Minor Junior Junior Computer Yes 7 min 5.03 sec 1 Edit Yes Neither satisfied nor dissatisfied 26—F Very satisfied
22 Excision of lesion on left lower abdomen. Minor Junior Junior Computer Yes 4 min 4.05 sec 2 Edits Yes Satisfied 70—M Satisfied
23 Excision of lesion on left chest. Minor Junior Junior Computer Yes 5 min 4.17 sec 2 Edits Yes Satisfied 79—M Very satisfied
24 Excision of lesion right knee. Minor Junior Junior Computer Yes 4 min 4.1 sec 2 Edits Yes Satisfied 70—M Dissatisfied
25 Excision of lesions on right groin and right hand. Minor Junior Junior Computer Yes 4 min 5.6 sec 3 Edits Yes Neither satisfied nor dissatisfied 52—M Satisfied
26 Excision of lesion on left breast. Minor Junior Junior Computer Yes 4 min 5.25 sec 2 Edits Yes Satisfied 81—F Very satisfied
27 Punch biopsy left cheek. Minor Junior Junior Computer Yes 4 min 4.0 sec 2 Edits Yes Satisfied 51—M Satisfied
28 Left thumb nailbed repair. Minor Junior Junior Computer Yes 10 min 4.2 sec 1 Edit Yes Satisfied 64—F Very satisfied
29 Left thumb nailbed repair. Minor Junior Junior Computer Yes 4 minutes 4.0 seconds 1 Edit Yes Satisfied 50—F Satisfied
30 Left index finger FDP—FDS—RDN—RDA repair. Intermediate Senior Junior Computer Yes 5 min 7.02 sec 4 Edits Yes Satisfied 32—M Satisfied
Open in a new tab

Figure 1.

Figure 1

Open in a new tab

ChatGPT-generated operative note showing adherence with the current guidelines by incorporating the required data points for an accurate operative note.

Patients involved in this study were both men (n=14) and women (n=16) with an average age of 51.25 years. Most of the patients were either very satisfied or satisfied (n=26) with the use of AI to generate operation notes. Only a small number of patients (n=4) were either dissatisfied or had no opinion on the matter.

We could also see that most of the operating surgeons expressed high levels of satisfaction (n=25) with only 4 being neutral and 1 being dissatisfied. Overall, most of the feedback was positive.

Furthermore, the AI-generated notes were enhanced with realistic images created using AI, which helped to illustrate incisions and suture sites (Figure 2 and 3). This additional feature shows the potential for AI to add value to medical documentation and improve the quality of care provided to patients.

Figure 2.

Figure 2

Open in a new tab

AI-generated image of a left hand using DALL-E AI platform. A text prompt was used to generate the requested image. This can be used as an illustration in operative notes.

Figure 3.

Figure 3

Open in a new tab

Hyper-realistic AI-generated images of upper and lower cheek skin lesions generated by DALL-E AI platform. This can be used to enhance the operative note and act as a visual aid or to indicate incisions and suture lines.

Overall, the testing of the ChatGPT platform reported its ability to generate high-quality operation notes efficiently and accurately, potentially saving time and improving the documentation process for medical professionals.

Discussion

The importance of accurate and thorough surgical operation notes extends beyond maintaining medical records and facilitating communication among health care providers for ongoing patient care. In addition to their medical significance, the quality of surgical notes can have notable legal and economic consequences. Meticulous record-keeping can facilitate audits and research, which can ultimately result in improved patient care. However, the evidence indicates that handwritten operative notes are associated with notable deficiencies that can have a negative effect on patient care3 and despite the availability of more advanced alternatives, many institutions in resource poor countries or publicly funded health care systems continue to rely on handwritten operative notes as their primary method for documenting surgical procedures. Although electronic health record systems have recently surfaced, they have limitations in terms of offering predetermined templates that require manual filling and lack intelligent features such as self-development or pattern recognition. Furthermore, their slow adoption rate and steep learning curve can be challenging for many users, and they may suffer from fragmentation and slow processing speeds.

Another feasible solution that has shown effectiveness involves creating an operative note proforma or template4,5 using a text processing software, which can assist surgeons in completing the necessary fields to adhere to the current guideline.6 Although effective, this solution can be time-consuming because it requires a desktop machine and requires the surgeon to create, fill, store, and edit the document from scratch, which can be a cumbersome process. In addition, it requires multiple data input points and actions in accordance with the current guidelines, as mentioned further:

  • 1.

    Date and time

  • 2.

    Names of the operating surgeon and assistant

  • 3.

    The operative procedure carried out

  • 4.

    Incision

  • 5.

    Operation diagnosis and the operative findings

  • 6.

    Details of the closure

  • 7.

    Any problems or complications

  • 8.

    Postoperative care instructions

  • 9.

    Estimated blood loss

  • 10.

    Details of tissue removed, added, or altered

  • 11.

    Signature

  • 12.

    Name of the theater anesthetist

LLMs are providing a cheap, universal, more powerful, and easy to use option that allows AI to analyze previous similar data sets and to find similarities in previously provided texts and apply them in a new context.7 A LLM is a powerful tool that reported efficacy in analyzing huge text data sets and finding solutions through processing massive databases,8 and we believe this can be a cost effective solution in publicly funded health systems or underfunded systems like the National Health Service in the United Kingdom.

Although we have tested ChatGPT on a limited number of procedures, we believe that there is a huge room for advancing its use into more complex procedures by fine tuning because one of the qualities of LLMs is that they are trainable and refinable. Refining a LLM involves modifying and tailoring a pre-existing model to execute specialized tasks or to better accommodate a specific domain.9 This procedure typically involves further training the model on a smaller, focused dataset that is pertinent to the intended task or topic, and this can be easily done by feeding the LLM a dataset of operations.10,11 The generated notes may require additional edits by medical professionals to ensure accuracy and completeness.

A recent study found that a LLM used by Bloomberg L.P. could process financial data to generate coherent text inputs for headlines and could search through economic data sets to provide insight on current and future trends12 another study found out that LLMs can react to common economic scenarios and it behaves in ways similar enough to humans.13

We have also noted that the fine tuning of LLMs can perform well in medically related tasks like answering patient questions using the Chat Doctor LLM, which can scrape data in real time and present them to patients as a reliable answer.14 LLMs also have shown high accuracy and speed in radiology report summaries.15

It is worth mentioning that the text prompt engineering plays a crucial role in using ChatGPT for generating operative notes because it helps in directing the model to produce relevant and specific content. By carefully crafting prompts that include essential keywords, context, and instructions, the model can be guided to generate more accurate and detailed operative notes. This process may involve iterative refinement and testing of various prompts to discover the most effective phrasing that reports desired results. In our experience, we found that mentioning the existing name, excision margin, and type of suture used in the prompt decreased the number of edits needed to complete the note. We expect that combining prompt engineering with fine tuning on surgery-specific datasets will ensure that the generated operative notes are not only tailored to individual procedures but also adhere to the required standards of medical documentation.

Despite the need to be edited, the overall process of editing operation notes generated by ChatGPT remained faster than the traditional method of handwriting or computer typing because the edits were largely less than 5 words and took ∼1-2 minutes on average to edit.

The edits could be categorized generally as the following: (1) Correcting specific medical terms or abbreviations or changing them, as in changing the word Confirmed to Suspected or adding a single surgical step (Mainly the WHO check list); (2) Adjusting numerical values, such as amount or type of local anesthetic; (3) Adding patient demographic characteristics or personal information; and (4) Updating names of involved medical staff.

The power of LLMs is not limited to generating text, but it can also perform more complex tasks like solving questions and providing reliable answers thanks to their data processing capabilities. For example, GPT-4 that was released in March 2023 outperformed ChatGPT that was released in November 2022 in solving neurosurgery oral board surgery questions achieving a score of 82.6%, whereas the 4-month-old model scored 62.4%.16 This shows that the advancements made in the field of AI are substantial, and there has been an exponential increase in research and publications in this area.

We could determine the benefits of using ChatGPT to generate operative notes to be mentioned in the following:

  • 1.

    Efficiency: ChatGPT could quickly generate accurate guideline-adhering operative notes in record time, saving health care professionals valuable time and allowing them to focus on advancing patient care.

  • 2.

    Consistency: ChatGPT showed impressive consistency in the formatting and content of operative notes, reducing the risk of errors and misinterpretations that can face operating surgeons after long procedures and exhausting days.

  • 3.

    Customizability: ChatGPT can be customized to generate notes specific to different procedures or surgeons, ensuring that the notes are tailored to the unique needs of the surgeon professional and their patients. We aim to explore this specific ability in future research.

  • 4.

    Accessibility: ChatGPT can be accessed from any device with an internet connection without the need for a computer, making it easy for health care professionals to generate notes on-the-go if needed.

  • 5.

    Integration: ChatGPT can be integrated with the current electronic health record systems if needed, making it easy to add generated notes directly to the patient records.

  • 6.

    Data analyzing: ChatGPT showed great potential in analyzing previous data and can be leveraged to suggest solutions, imaging, laboratory tests, and postoperative plans on the basis of similar previous patient results and symptoms, which can aid research and greatly improve care.

Current Limitations

Despite being recently introduced to the market, the ChatGPT platform has already gained attentionin the medical documentation domain but is still facing lots of challenges, such as personalization of documents that need human input.17 It is worth mentioning that the speed of note generation by the online platform might be variable depending on the version of Chat GPT used and the speed of the internet connection but in all cases, it was limited to only seconds difference, and it has proven better speed in creating operation notes over human written notes. One of the current limitations of ChatGPT in formulation of medical notes is the limited context awareness as currently without full access to patient file notes and data, ChatGPT may not be able to comprehend the full context of the medical notes, including the patient’s previous medical history, previous treatments, or specific details of the surgical procedure.

There are also some legal and ethical concerns surrounding the use of AI in health care, such as data privacy and the protection of patient information, but we think our current use of ChatGPT is not conflicting with the current General Data Protection Regulation and data governance rules because no patient information was added to the platform and only surgical steps are generated but care should be given if further implementation of the platform into patient care process because this will give access to the software to patient sensitive data and might cause a confidentiality risk and negatively affect patients’ trust in care providers.18

Moreover, during our survey we noted that some older patients and surgeons were reluctant toward expressing satisfaction or optimism toward the new technology, which might be a limitation toward further implementation of LLMs into the medical sector because some health care providers may be hesitant to adopt a new technology, or find the use of ChatGPT or other AI tools to be too complex or time-consuming and prefer to work with traditional methods. Finally, integrating ChatGPT and similar tools into the current IT infrastructure may require technical resources and expertise and may pose compatibility issues or other logistical challenges.

Conclusion

ChatGPT’s recently launched beta version has reported operation note efficiency and accuracy, as shown by this study. We believe that it has the potential to be a highly beneficial AI language modeling tool for modern plastic surgeons. By using ChatGPT, surgeons can save time, increase their knowledge of surgical procedures, and generate customizable, efficient, and safe operation notes that comply with current guidelines. Despite being a relatively new tool, ChatGPT has shown promise in its ability to assist health care providers in a variety of contexts.

Potential Competing Interests

The authors report no competing interests

Supplemental Online Material

Video

Video showing the process of note generation by Chat GPT by entering a text command to the field and clarifying the name of procedure.

Download video file (19.7MB, mp4)

References

  • 1.Singh R., Chauhan R., Anwar S. Improving the quality of general surgical operation notes in accordance with the Royal College of Surgeons guidelines: a prospective completed audit loop study. J Eval Clin Pract. 2012;18(3):578–580. doi: 10.1111/J.1365-2753.2010.01626.X. [DOI] [PubMed] [Google Scholar]
  • 2.Optimizing language models for dialogue. ChatGPT. https://openai.com/blog/chatgpt/
  • 3.Nzenza T.C., Manning T., Ngweso S., et al. Quality of handwritten surgical operative notes from surgical trainees: a noteworthy issue. ANZ J Surg. 2019;89(3):176–179. doi: 10.1111/ANS.14239. [DOI] [PubMed] [Google Scholar]
  • 4.Parwaiz H., Perera R., Creamer J., Macdonald H., Hunter I. Improving documentation in surgical operation notes. Br J Hosp Med (Lond) 2017;78(2):104–107. doi: 10.12968/hmed.2017.78.2.104. [DOI] [PubMed] [Google Scholar]
  • 5.Chan B.K.Y., Exarchou K., Corbett H.J., Turnock R.R. The impact of an operative note pro0forma at a paediatric surgical centre. J Eval Clin Pract. 2015;21(1):74–78. doi: 10.1111/jep.12242. [DOI] [PubMed] [Google Scholar]
  • 6.Mustafa M.K.E., Khairy A.M.M., Ahmed A.B.E. Assessing the quality of orthopaedic operation notes in accordance with the Royal College of Surgeons guidelines: an audit cycle. Cureus. 2020;12(8) doi: 10.7759/cureus.9707. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Kim G., Baldi P., McAleer S. Language models can solve computer tasks. Preprint. Published online March. 2023;30 doi: 10.48550/arXiv.2303.174918. arXiv:2303.17491. [DOI] [Google Scholar]
  • 8.Liu Y, Han T, Ma S, et al. Summary of ChatGPT/GPT-4 research and perspective towards the future of large language models. Preprint. Published online April 4, 2023. arXiv:2304.01852. doi: 10.48550/arXiv.2304.01852
  • 9.Wu C, Zhang X, Zhang Y, Wang Y, Xie W. PMC-LLaMA: further finetuning LLaMA on medical papers. Preprint. Published online April 27, 2023. arXiv:2304.14454. doi: 10.48550/arXiv.2304.14454
  • 10.Fine-tuning—OpenAI API. OpenAI. https://platform.openai.com/docs/guides/fine-tuning
  • 11.Boiko DA, MacKnight R, Gomes G. Emergent autonomous scientific research capabilities of large language models. Preprint. Published online April 11, 2023. arXiv:2304.05332. doi: 10.48550/arXiv.2304.05332
  • 12.Wu S, Irsoy O˙, Lu S, et al. BloombergGPT: a large language model for finance. Preprint. Published online March 30, 2023. arXiv:2302.17564. doi: 10.48550/arXiv.2303.17564
  • 13.Horton JJ. Large language models as simulated economic agents: what can we learn from Homo Silicus? Preprint. Posted online April 10. arXiv 2301.075432023. doi: 10.3386/W31122
  • 14.Li Y., Li Z., Zhang K., Dan R., Zhang Y. ChatDoctor: a medical chat model fine-tuned on LLaMA model using medical domain knowledge. Preprint. Published online March. 2023;24 doi: 10.48550/arXiv.2303.14070. arXiv:2303.14070. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Ma C, Wu Z, Wang J, et al. ImpressionGPT: An Iterative Optimizing Framework for Radiology Report Summarization with ChatGPT. Preprint. Published online April 17, 2023. arXiv:2304.08448. doi: 10.48550/arXiv.2304.08448
  • 16.Ali R, Tang OY, Connolly ID, et al. Performance of ChatGPT, GPT-4, and Google Bard on a Neurosurgery Oral Boards Preparation Question Bank. Preprint. Posted online April 12, 2023. medRxiv:2023.04.06.23288265. doi:10.1101/2023.04.06.23288265 [DOI] [PubMed]
  • 17.Patel S.B., Lam K. ChatGPT: the future of discharge summaries? Lancet Digit Health. 2023;5(3):e107–e108. doi: 10.1016/S2589-7500(23)00021-3. [DOI] [PubMed] [Google Scholar]
  • 18.Powles J., Hodson H. Google DeepMind and healthcare in an age of algorithms. Health Technol (Berl) 2017;7(4):351–367. doi: 10.1007/S12553-017-0179-1. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Video

Video showing the process of note generation by Chat GPT by entering a text command to the field and clarifying the name of procedure.

Download video file (19.7MB, mp4)

Articles from Mayo Clinic Proceedings: Digital Health are provided here courtesy of Elsevier

RESOURCES