Background: Professor Fisher’s legacy, defined by meticulous observation, curiosity, and profound knowledge, has established a foundational cornerstone in medical practice. However, the advent of automated algorithms and artificial intelligence (AI) in medicine raises questions about the applicability of Fisher’s principles in this era. Our objective was to propose adaptations to these enduring rules, addressing the challenges and leveraging the opportunities presented by digital health. Summary: The adapted rules we propose advocate for the harmonious integration of traditional bedside manners with contemporary technological advancements. The judicious use of advanced devices for patient examination, recording, and sharing, while upholding patient confidentiality, is pivotal in modern practice and academic research. Additionally, the strategic employment of AI tools at the bedside, to aid in diagnosis and hypothesis generation, underscores their role as valued complements to clinical reasoning. These adapted rules emphasize the importance of continual learning from experience, literature, and colleagues, and stress the necessity for a critical approach toward AI-derived information, which further consolidates clinical skills. These aspects underscore the perpetual relevance of Professor Fisher’s rules, advocating not for their replacement but for their evolution. Thus, a balanced methodology that adeptly utilizes the strengths of AI and digital tools, while steadfastly maintaining the core humanistic values, arises as essential in the modern practice of medicine. Key Messages: A commitment between traditional medical wisdom and modern technological capabilities may enhance medical practice and patient care. This represents the future of medicine – a resolute commitment to progress and technology, while preserving the essence of medical humanities.

The story of Professor Fisher is an indispensable chapter for any physician, be they neurologists or otherwise [1‒4]. Rules and lessons that extend beyond the scientific realm elevate his narrative into a true compendium, particularly beneficial for beginners [3]. His character, marked by meticulous observation, his curiosity [1‒3], and his profound knowledge of the anatomical and pathophysiological foundations of Neurology, led him to describe or elucidate conditions such as transient global amnesia, carotid dissection, atrial fibrillation as a risk factor for stroke, hemorrhagic infarction, lacunar syndromes, normal pressure hydrocephalus, thalamic syndromes, and various other conditions, including the Miller Fisher Syndrome, in his legacy to Medicine [4‒10].

Becoming a doctor today involves venturing into the realms of data science, automated algorithms, and artificial intelligence (AI). These fields dominate the news and are now pervasive in hospitals and clinics. Yet, with every publication in AI, with each successful step in creating a new algorithm, we think about how Professor Fisher would handle these; and then numerous questions arise: How would Professor Fisher’s rules cope with such technologically advanced medicine today? Would Professor Fisher’s rules remain intact amidst the deluge of applications and algorithms touted as superior to the human mind? In 1982, Caplan published a monologue on the principal rules of Professor Fisher [3]. Twelve years after Professor Fisher’s demise and 32 years following the publication [3], we conjecture what the changes or adaptations to Professor Fisher’s rules might be in an era engulfed by digital health [11]. If not aptly managed, this digital shift could create chasms between doctors and patients, between humanism and technique, between creativity and stagnation, between learning and mere repetition of actions [12].

Digital health encompasses the integration of technology with healthcare practices to enhance patient care, involving tools such as advanced electronic health records and telemedicine [11]. Big data refer to the vast quantities of health-related data generated from multiple sources, crucial for insightful analytics. Automated algorithms are sets of rules or procedures coded into software to perform tasks without human intervention, often used in analyzing complex medical data. Decision-support platforms are systems that help clinicians make better treatment decisions by providing data-driven insights. AI in health refers to the use of machine learning and other AI techniques to mimic human cognition in the analysis, interpretation, and comprehension of complex medical data [11]. These emerging concepts are becoming indispensable in clinical practice, significantly transforming how healthcare is delivered and managed – while also raising concerns about professional independence, equality, inclusion, and humanism [11, 12].

Professor Fisher always conducted himself with humility, open to learning – from patients, students, residents, or teachers. Thus, his lesson of humility toward knowledge would probably welcome assistance from digital resources. Generative AI tools, which enable the comprehension and understanding of free discourse [11‒13], would be endorsed by him as a complement to bedside observations and learning, for instance.

We humbly suggest adaptations of Professor Fisher’s rules [3] to the current landscape of Medicine. Figure 1 summarizes these adaptations. We fervently believe that it is possible to meld the best of both worlds – and then preserve, intact, the essence of Fisher’s rules that will never wane with time.

At the Bedside, Your Laboratory Awaits; Study the Patient with Solemn Intent

Spend time alongside your patient. Sit close [14], engage in conversation, and examine the patient. With the patient’s consent, record photos and videos using a digital camera. Optical devices, with their intricate craftsmanship, can capture the retina in high definition, seamlessly exporting these images directly to specialists. Moreover, algorithms of AI stand ready to suggest modifications, even at the patient’s bedside, with a blend of precision and grace [15‒17]. This material, while preserving the patient’s sensitive data, can be invaluable for teaching other doctors or seeking opinions from experts even in different parts of the world. Record your hypotheses. Make summaries of the clinical history and findings from the physical examination – generative AI tools can assist in contemplating alternative diagnostic hypotheses [15]. Participating in digital forums can also aid in generating hypotheses. Sharing clinical insights is a helpful exercise for making diagnoses more accurately. Following the rules of privacy strictly is crucial – the mentioned forums are useful for sharing knowledge and clinical tips, not for sharing patient data.

Resolve Bedside Matters as They Emerge

Resolving bedside situations presents an opportunity to dispel uncertainties. Beyond the clinical history, observation, and meticulous neurological examination, algorithms processing serial physiological data, portable apps predicting the likelihood of clinical outcomes and complications, and AI software identifying findings in imaging studies can assist, more swiftly, in consolidating or perhaps even refuting initial hypotheses right at the bedside [16]. We suppose this arsenal enables us to provide quicker, more comprehensive, and possibly more accurate responses – yet, it is vital to remember that diagnostic tests are, indeed, merely complementary to clinical reasoning. Do not hesitate to graph the patient’s evolution. Creating something like a mind map, a dashboard, and a timeline can help clarify the phenomena that occurred with the patient, as well as better organize ideas. Tools based on generative AI can organize medical notes into mind maps and timelines – free tools such as ChatMind®, Taskade®, and Getmind® may be useful.

Formulate a Hypothesis, and Then Rigorously Seek Its Refutation or Exception before Deeming It Valid

Before asserting an idea decisively, think and rethink. Consider ways to validate your idea. International networks of collaboration, especially involving large, mixed populations in low- and middle-income countries, and decentralized clinical trials [17] are resources to test hypotheses and generalizations more accurately [17]. The ease of data sharing and access, the big data, allows us to go beyond our walls. These resources enable simultaneous personalization and generalization of results, ensuring our ideas or hypotheses remain robust over time [18]. Simultaneously, all sensitive data should be included solely in well-designed and protected datasets, in accordance with privacy requirements. The AI can significantly enhance the capability of conducting meta-analyses, particularly when dealing with complex datasets. AI-driven tools (such as RobotReviewer™, Silvi.ai™, ExaCT™, and Rayyan™) can automate the extraction and synthesis of relevant data, assess the quality of studies, and handle heterogeneous data types effectively. This process not only speeds up the meta-analysis but also enhances its accuracy and reliability. Adopting common data elements, like those from the NINDS Common Data Elements initiative, can improve data sharing, standardize quality metrics, and facilitate participation in large clinical trials or cohorts in the field of Vascular Neurology.

Engage Ceaselessly in Projects; They Render Daily Life More Profound

When a hypothesis blooms, data collection at the bedside and in the literature must be undertaken. Data mining tools powered by AI assist in processing natural language [18‒22], extracting data from laboratory and neuroimaging tests, as well as accessing scientific publications more quickly and concisely. This enables the optimization of time that would have been consumed in these previously manual activities, thereby increasing productivity. Accessing international databases of healthy individuals allows for the creation of a “control,” facilitating comparisons with pathological cases observed in the clinic [18‒21]. Maintaining and increasing contact with diverse patients, even those deemed healthy enables physicians to identify more patterns and, consequently, make better decisions.

In Diagnosing, Consider the Quintet of Common Signs – Historical, Physical, or Laboratory-Based

When three of these findings are absent, in each realm, your hypothesis likely errs. This threshold, however, may shift as years unfurl. Be vigilant and acknowledge that diagnostic models can now eclipse traditional diagnostic criteria – oft defined by mere consensus of experts. With the vast sharing of data, models born of big data’s embrace can distill common clinical and laboratory traits [19, 20]. Incorporate findings born of genomics to support or refuse your hypothesis [23]. Furthermore, AI could standardize the definitions of clinical events and diagnostic criteria. This uniformity would facilitate the full implementation of Professor Fisher’s strategy both within and across many healthcare sites, as well as longitudinally within individual patients.

Quantify with Precision

Whether in spoken or written word, report with detail what you see. Scales and scores, published and validated in the literature, aid in rendering communication more objective. Embark these tools upon platforms of digital records or employ calculators in mobile applications [18] – keep these resources at hand. These metrics assist in the comparison of the physical examination, for instance, as the months pass by. The recording of images, video, and even the patient’s voice, when consented and anonymized, can be invaluable for such intricacy [19‒21]. Maintaining the privacy of sensitive data is an inherent act in the practice of any health professional. However, the increasing digitalization of health requires specific actions in this regard, such as maintaining cloud-based servers ideally located within the institution, blockchain resources throughout the information journey, and caution when inputting data into generative models. For instance, employing a generative AI system not designed for healthcare purposes to create clinical notes could inadvertently lead to a clinician relaying sensitive information to an insecure web service. Similarly, using a computer program intended for other purposes to identify skin lesions could also present security risks. Other AI-based imaging diagnostic tools also raise privacy concerns. Regulatory norms for the use of these data are still rare but measures to obtain broad patient consents or anonymization models could be helpful to prevent any possibility of identifying individuals. These measures include pseudonymization, stripping datasets of patients’ identifying variables according to the HIPAA 18 items “Safe Harbor” method (US), and anonymization. Using AI-based platforms that securely capture the sounds of medical consultations, autonomously write clinical notes, categorize findings, summarize cases, and integrate with data from other electronic health records may be a wise idea to save time and enhance the reporting of this information [19‒21].

Case Nuances Elevate the Expert above the Journeyman

Every nuance in the clinical history and physical examination holds paramount importance. Over time, we learn to render finer diagnoses by echoing patterns witnessed – it is the melding of observed patterns in our clinical practice with insights gleaned from the literature. AI can reliably detect and characterize nuances, highlight their occurrences, and systematically record these insights in specialized databases, such as the O.S.L.E.R.™ and Freed™ tools. No matter how advanced and accurate the clinical decision-support tools powered by AI may be, the physician’s acumen remains crucial in interpreting the outcomes of algorithms [24‒27]. The physician continues to be the gold standard in clinical diagnosis.

Gather and Classify Phenomena; Their Essence and Mechanism May Reveal Themselves with Sufficient Accumulation

Documenting uncommon findings is essential for discerning patterns and, perhaps, describing new clinical conditions. Text editors embedded in tablets and cloud environments allow for the recording of notes, and attaching photographs, videos, and neuroimaging results, all labeled with keywords. Cloud-based AI tools enable the creation of descriptive texts for findings in photographs [18‒21]. Masks derived from natural language processing/large language algorithms facilitate the identification of similar clinical descriptions in local databases or those of other institutions, drawn from data lake – healthcare data lakes are specialized data repositories that allow health systems and payer organizations to store, access, and analyze extensive amounts of information centrally. Multicentrically updated databases now catalog clinical findings, as part of an effort to standardize these observations into categories [28] (Online Mendelian Inheritance in Man, OMIM®, for instance). The OMIM® is a comprehensive and authoritative compendium of human genes and genetic phenotypes. Freely accessible and updated daily, OMIM provides full-text, referenced overviews of all known Mendelian disorders and over 16,000 genes. It emphasizes the relationship between phenotype and genotype, offering extensive links to additional genetic resources. The entries in OMIM are authored and edited at the McKusick-Nathans Institute of Genetic Medicine, Johns Hopkins University School of Medicine. To record and access these findings, whether in one’s database or through multicentric sharing, is a step toward a more enlightened and interconnected realm of medical science [17].

Embrace Only That Which You Have Personally Affirmed

Be skeptical of what you hear or read – especially when such information comes from generative AI, despite the increasingly growing reliability of these tools. Regarding these tools, trust but always verify the outputs [20]. Always check the references, review the authors’ backgrounds, and verify the physiological nexus behind an explanation or statistical result. With the consolidation of social networks and instant messaging, unfortunately, false information spreads easily. Although somewhat nebulous and occasionally deemed outdated, Web 2.0 refers to extensive user-generated content, where users play a central role in creating, updating, and disseminating content. The Web 2.0 allowed user protagonism, but at the risk of fake news in the health field. Always verify the information before passing it on. Reliable journals and experts should be consulted [29‒31].

Glean Wisdom from Personal Encounters and the Seasoned Insights of Literature and Peers

Gaining knowledge is fundamental to growth. Both physical books and digital resources should continuously serve as sources of enlightenment, embodying the power to leverage wisdom. The guiding principle is to learn, reason, and refine. Contemporary platforms for ongoing medical education now embrace AI tools to discern the learner’s gaps, crafting a personalized path of intellectual formation [31]. Generative AI instruments possess the prowess to condense scientific articles and even juxtapose disparate texts, thereby optimizing scholarly reviews. Moreover, virtual and/or augmented reality resources may improve educational objectives, immersing students, residents and even specialists in a more realistic and adjustable field of study.

Didactic Dialogs Enrich the Lecturer; We Best Instruct through Listening, Inquiry, and Demonstration

Teaching serves as a vital conduit for learning. Listening to feedback and assessing the knowledge acquired by the audience are methods to pinpoint areas needing refinement in the lesson. AI-based tools can assist in creating slides with educational features, intuitive infographics, and even instructional videos. Generative AI can produce texts tailored to the target audience’s language [20‒22]. Tools that generate live quizzes facilitate real-time assessment of the audience’s learning, also including downloadable mobile applications associated with the lecture content and links to educational content, and creating dashboards that display performance metrics for the content presented in the class. For example, such tools include LessonPlans.ai™, Learnt.ai™, Teachology.ai™, and Jasper™.

Write Diligently; Let Your Work Enlighten Others

Set forth objectives for your scholarly publications. Employ tools that cluster and arrange bibliographic references. The aids of generative AI can assist in penning case reports and discussions – yet, it is imperative to use these instruments as ancillary resources, not as definitive solutions, ethically declaring their adoption. When aptly utilized, these resources augment your productivity – and assuredly, you shall surpass your expectations. Furthermore, tools that automate the evaluation of movement disorders (such as PDMonitor®), skin alterations (such as First Derm™), and AI tools applied to the bedside identification of retinal changes (such as Eyer Maps™), as well as data mining in clinical notes for patient identification (such as Nabla Copilot™, Epic®, Prognos™, and Zeus™) are used for the publication of case series.

Attend Meticulously to the Nuances of a Diagnosed Patient; Such Vigilance Enlightens Future Encounters with Enigmatic Cases; Eschew Blind Acquiescence to Any Diagnosis

Never cease to investigate (whether through complementary exams or by delving deeper into the physical examination in search of new findings) when a diagnosis is made. And even if the diagnosis is correct, do not be satisfied merely by making the diagnosis [18]. To listen, to note, and to study the description of a phenomenon is an enriching exercise. Therefore, no matter how advanced the automated resources of clinical decision support and/or diagnostic platforms may be, do not blindly trust the suggested diagnoses (always consider them as possible diagnoses, not definitive ones), and delve into the intricacies of the diagnosis. Examine the methodology by which the algorithm generates its suggested diagnoses [27]. However, this analysis may frequently be unfeasible, as numerous algorithms are based on deep neural network architectures, which render the intermediate processes of the model opaque.

Heed the Novice’s Voice, for Wisdom Often Whispers

Knowledge today is manifold and accessible. Even novice students may access entire works, summaries, and generative AI tools that can compile, research, and encapsulate the most advanced knowledge [19]. Furthermore, lectures and conferences can be broadcast in real-time across the globe, enhancing the access and exchange of information between students and experts. Forums and social networks now enable the reception of queries, concerns, and insights from people worldwide. This deluge of ideas, a veritable perennial brainstorm, allows for the emergence of thoughts and insights from all directions, even from those who are just beginning their academic journey. For instance, the Delphi method is a structured approach to gathering information from experts and contributors. Web-based platforms (such as eDelphi®) were designed for qualitative use, it emphasizes expert engagement and discussion, enabling the Delphi process to not only collect existing information but also generate new insights.

Avoid Prematurely Categorizing a Case or Disorder into an Ill-Fitting Diagnostic Category

As with the previous rule, do not be content with first-glance diagnoses. Even with resources supporting diagnosis, do not trust blindly. Remember, these algorithms depend on human data input – opening the door to errors in recognizing signs and describing symptoms. Thus, leaving a diagnosis unknown is not a matter of shame; on the contrary, it stimulates the desire to continue examining and reevaluating the patient continuously. As accurate as it may be in various scenarios, AI skirts the limitations of medical knowledge and is confined by its inherent limitations [32]. Syndromic clustering in neurological clinical practice involves grouping patients based on shared symptoms and characteristics, which can be significantly enhanced by machine learning algorithms. Machine learning offers a data-driven approach to identify patterns and correlations within complex neurological data [20].

The Patient Is Always Doing the Best He Can

Support the patient and their family. Be humane, sensitive, present, empathetic, kind, and dedicated. Place your chair beside the patient’s bed and be patient. Do not reduce your medical practice to scores, predictive models, decision-support algorithms, or complementary tests. Being a doctor surpasses all the accuracy and power of AI [33, 34]. When AI becomes more empathetic than the doctor, something is profoundly amiss. We care for people; we deal with people; be human as well. In addition, be able to adopt digital supports to enhance health communication between doctors and their patients. Generative AI tools can act as support chats for emotional support and encouragement of physical activity and other post-stroke rehabilitation measures, in addition to medication reminders, dietary guidance, family cohesion, fall detection, social interaction, and even alerts for caregivers. Popular home assistants, such as Alexa®, already feature some of these capabilities.

Maintain a Lively Interest in Patients as People

Our patients cannot be solely represented by even the most complex algorithms or scales [33]. Even when broadly utilized to categorize patients based on prior functionality, it is essential to see beyond rigid data; and consider the limitations of algorithms in perceiving the social and psychological dimensions of people. AI and related technologies, such as ecological momentary assessment, actigraphy, pedometers, implantable EKG recorders, and ambulatory EEG, allow for continuous monitoring of patients outside the clinic – such as NeuroBrave™, BioButton® and Apollo Labs™. These tools help capture the broader context of a patient’s life, providing insights into their health and personal experiences beyond brief clinical interactions. While many attempt to quantify health variables using algorithms, it is crucial to approach such automated results with skepticism. A deep understanding of medicine is essential to determine the best care plan for each patient, rather than relying solely on algorithmic outputs [34]. In summary, understanding the patients’ work environment, their families, where they live, spirituality, and other aspects that contribute to their uniqueness is essential. It is this sense of humanism, inclusion, diversity, and respect for the human person that keeps us more robust than any AI to date.

As healthcare ventures into the realms of AI and data science, adapting Fisher’s rules becomes imperative to maintain the essence of clinical excellence. Beyond the well-established concept of clinical-radiological correlation, we are embracing the emerging concept of “clinical-AI correlation.” At this stage of growth and consolidation of AI in Medicine, significant clinical expertise is required to assess the results from these algorithms. It is here that Professor Fisher’s rules become unquestionably valuable. Simultaneously, the data sources that inform and train AI algorithms must be critically evaluated, considering variations among examiners and methods of data collection. By integrating advanced digital tools with the core values of observation, humility, continuous learning, and patient-centered care, we proposed a modern reinterpretation of Fisher’s legacy. This adaptation not only preserves but also enhances the physician-patient relationship amidst technological advancements. This essential relationship also extends to the neurological exam – so acclaimed and meticulously conducted by Professor Fisher – in which new devices such as portable retinographers, wearables, and sensors for detecting changes in skin, gait, and movements are transforming the way we perform neuro exams. These tools can enhance and facilitate the examination, but they should never supplant the neurologist’s skills in observation, interpretation, and integration. Finally, the Prof. Fisher’s Rules adaptation underscores the irreplaceable value of human insight, empathy, and critical thinking in the face of automated diagnostics and algorithm-based decision-making.

I extend my heartfelt gratitude to Professor Jay P. Mohr for his encouragement and the esteemed manner in which he imparted the teachings of Professor Fisher to me. His support has been invaluable in my academic journey. Additionally, I would like to express my profound appreciation to Professor Louis Caplan for his kind and adept review and suggestions throughout the manuscript. His insights and expertise have significantly enhanced the quality of this work.

Declaration of generative AI and AI-assisted technologies in the writing process: during the preparation of this work, the authors used CHATGPT 4 in order to review grammar and language. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication. CHATGPT® 4.0 was also employed to translate Portuguese into English. CHATMIND® was utilized for summarizing the rules into a smart graph. No other generative AI was used for text production in this process. The authors have read and approved the submitted manuscript. The manuscript has not been submitted or published elsewhere, either in whole or in part. All authors grant the right to publish any and all data from this research. This is a review paper. No data from human subjects were included in this paper.

The authors declare that there is no conflict of interest related to the study design, execution and analysis, and manuscript conception, planning, writing and decision to publish.

This study was not supported by any sponsor or funder.

Andrade, J.B.C.: design, review, data collection, and contact with mentors (J.P. Mohr and L. Caplan); Mendes, G.N.N.: review and final version; and Silva, G.S.: design and final version.

1.
Caplan
LR
,
Mohr
JP
,
Ackerman
RH
.
In memoriam: charles miller Fisher, MD (1913-2012)
.
Arch Neurol
.
2012
;
69
(
9
):
1208
9
.
2.
Mohr
JP
,
Caplan
LR
,
Kistler
JP
,
Memoriam
C
.
C. Miller Fisher: an appreciation
.
Stroke
.
2012
;
43
(
7
):
1739
40
.
3.
Caplan
LR
.
Fishers rules
.
Arch Neurol
.
1982
;
39
(
7
):
389
90
.
4.
Caplan
L
.
Caplan-Fisher rules
.
Stroke
.
2021
;
52
(
5
):
e155
9
.
5.
Caplan
LR
.
The last 50 years of cerebrovascular disease: Part 1
.
Int J Stroke
.
2006
;
1
(
2
):
104
8
.
6.
Fisher
CM
.
The history of cerebral embolism and hemorrhagic infarction
. In:
Furlan
A
, editor.
The heart and stroke
.
Berlin
:
Springer-Verlag
;
1987
.
Vol. 1
. p.
3
16
.
7.
Fisher
CM
.
Lacunes: small deep cerebral infarcts
.
Neurology
.
1965
;
15
:
774
84
.
8.
Fisher
CM
,
Caplan
LR
.
Basilar artery branch occlusion: a cause of pontine infarction
.
Neurology
.
1971
;
21
(
9
):
900
5
.
9.
Mohr
JP
.
Historical perspective
.
Neurology
.
2004
;
62
(
8_Suppl l_6
):
83
6
.
10.
Mohr
JP
.
Stroke data banks
.
Stroke
.
1986
;
17
:
171
2
. [editorial].
11.
Altunisik
E
.
Artificial intelligence and cerebrovascular diseases: ChatGPT Model
.
Cerebrovasc Dis
.
2023
;
53
(
3
):
354
8
.
12.
Ognjanovic
I
.
Artificial intelligence in healthcare
.
Stud Health Technol Inform
.
2020
;
274
:
189
205
.
13.
Lee
P
,
Bubeck
S
,
Petro
J
.
Benefits, limits, and risks of GPT-4 as an AI chatbot for medicine
.
N Engl J Med
.
2023
;
388
(
13
):
1233
9
.
14.
Iyer
R
,
Park
D
,
Kim
J
,
Newman
C
,
Young
A
,
Sumarsono
A
.
Effect of chair placement on physicians’ behavior and patients’ satisfaction: randomized deception trial
.
satisfaction: randomized deception trial BMJ
.
2023
;
383
:
e076309
.
15.
Johnson
SB
,
King
AJ
,
Warner
EL
,
Aneja
S
,
Kann
BH
,
Bylund
CL
.
Using ChatGPT to evaluate cancer myths and misconceptions: artificial intelligence and cancer information
.
JNCI Cancer Spectr
.
2023
;
7
(
2
):
015
.
16.
Jo
H
,
Kim
C
,
Gwon
D
,
Lee
J
,
Lee
J
,
Park
KM
, et al
.
Combining clinical and imaging data for predicting functional outcomes after acute ischemic stroke: an automated machine learning approach
.
Sci Rep
.
2023
;
13
(
1
):
16926
.
17.
Thomas
KA
,
Kidziński
Ł
.
Artificial intelligence can improve patients’ experience in decentralized clinical trials
.
Nat Med
.
2022
;
28
(
12
):
2462
3
.
18.
Chen
TC
,
Kaminski
E
,
Koduri
L
,
Singer
A
,
Singer
J
,
Couldwell
M
, et al
.
Chat GPT as a neuro-score calculator: analysis of a large language model’s performance on various neurological exam grading scales
.
World Neurosurg
.
2023
;
179
:
e342
7
.
19.
Gupta
A
.
3 predictions for AI in Healthcare in 2024
.
Google
;
2024
. Avaliable from: https://blog.google/technology/health/google-ai-and-health/3-predictions-for-ai-in-healthcare-in-2024/
20.
Haug
CJ
,
Drazen
JM
.
Artificial intelligence and machine learning in clinical medicine, 2023
.
N Engl J Med
.
2023
;
388
(
13
):
1201
8
.
21.
De Rosario
H
,
Pitarch-Corresa
S
,
Pedrosa
I
,
Vidal-Pedrós
M
,
de Otto-López
B
,
García-Mieres
H
, et al
.
Applications of natural language processing for the management of stroke disorders: scoping review
.
JMIR Med Inform
.
2023
;
11
:
e48693
.
22.
Dias
R
,
Torkamani
A
.
Artificial intelligence in clinical and genomic diagnostics
.
Genome Med
.
2019
;
11
(
1
):
70
.
23.
Herzog
L
,
Kook
L
,
Hamann
J
,
Globas
C
,
Heldner
MR
,
Seiffge
D
, et al
.
Deep learning versus neurologists: functional outcome prediction in LVO stroke patients undergoing mechanical thrombectomy
.
Stroke
.
2023
;
54
(
7
):
1761
9
.
24.
Alaka
SA
,
Menon
BK
,
Brobbey
A
,
Williamson
T
,
Goyal
M
,
Demchuk
AM
, et al
.
Functional outcome prediction in ischemic stroke: a comparison of machine learning algorithms and regression models
.
Front Neurol
.
2020
;
11
:
889
.
25.
Miller
MI
,
Orfanoudaki
A
,
Cronin
M
,
Saglam
H
,
So Yeon Kim
I
,
Balogun
O
, et al
.
Natural Language processing of radiology reports to detect complications of ischemic stroke
.
Neurocrit Care
.
2022
;
37
(
Suppl 2
):
291
302
.
26.
Sung
S
,
Chen
C
,
Pan
R
,
Hu
Y
,
Jeng
J
.
Natural language processing enhances prediction of functional outcome after acute ischemic stroke
.
J Am Heart Assoc
.
2021
;
10
(
24
):
e023486
.
27.
Moisset
X
,
Ciampi de Andrade
D
.
Neuro-ChatGPT? Potential threats and certain opportunities
.
Rev Neurol Paris
.
2023
;
179
(
6
):
517
9
.
28.
Yang
Y
,
Tang
L
,
Deng
Y
,
Li
X
,
Luo
A
,
Zhang
Z
, et al
.
The predictive performance of artificial intelligence on the outcome of stroke: a systematic review and meta-analysis
.
Front Neurosci
.
2023
;
17
:
1256592
.
29.
Dave
T
,
Athaluri
SA
,
Singh
S
.
ChatGPT in medicine: an overview of its applications, advantages, limitations, future prospects, and ethical considerations
.
Front Artif Intell
.
2023
;
6
:
1169595
.
30.
Day
T
.
A preliminary investigation of fake peer-reviewed citations andReferences generated by ChatGPT
.
The Prof Geographer
.
2023
;
75
(
6
):
1024
7
.
31.
Ayers
JW
,
Poliak
A
,
Dredze
M
,
Leas
EC
,
Zhu
Z
,
Kelley
JB
, et al
.
Comparing physician and artificial intelligence chatbot responses to patient questions posted to a public social media forum
.
JAMA Intern Med
.
2023
;
183
(
6
):
589
96
.
32.
Verghese
A
,
Shah
NH
,
Harrington
RA
.
What this computer needs is a physician: humanism and artificial intelligence
.
JAMA
.
2018
;
319
(
1
):
19
20
.
33.
Kalra
J
,
Rafid-Hamed
Z
,
Seitzinger
P
.
Artificial intelligence and humanistic medicine: a symbiosis
. In:
Kalra
J
,
Lightner
NJ
,
Taiar
R
, editors.
Advances in human factors and ergonomics in healthcare and medical devices. AHFE 2021. Lecture notes in networks and systems
.
Cham
:
Springer
;
2021
.
Vol. 263
. p.
3
8
.
34.
Ostherr
K
.
Artificial intelligence and medical humanities
.
J Med Humanit
.
2022
;
43
(
2
):
211
32
.