Natural language processing in clinical neuroscience and psychiatry: A review

Telepsychiatry and Artificial Intelligence: A Structured Review of Emerging Approaches to Accessible Psychiatric Care

BACKGROUND/OBJECTIVES

Artificial intelligence is rapidly permeating the field of psychiatry. It offers novel avenues for the diagnosis, treatment, and prediction of mental health disorders. This structured review aims to consolidate current approaches to the application of AI in telepsychiatry. In addition, it evaluates their technological maturity, clinical utility, and ethical-legal robustness.

METHODS

A systematic search was conducted across the PubMed, Scopus, and Google Scholar databases for the period spanning 2015 to 2025. The selection and analysis processes adhered to the PRISMA 2020 guidelines. The final synthesis included 44 publications, among which 14 were empirical studies encompassing a broad spectrum of algorithmic approaches-ranging from neural networks and natural language processing (NLP) to multimodal architectures.

RESULTS

The review revealed a wide array of AI applications in telepsychiatry, encompassing automated diagnostics, therapeutic support, predictive modeling, and risk stratification. The most actively employed techniques include natural language and speech processing, multimodal analysis, and advanced forecasting models. However, significant barriers to implementation persist-ethical (threats to autonomy and risks of algorithmic bias), technological (limited generalizability and a lack of explainability), and legal (ambiguous accountability and weak regulatory frameworks).

CONCLUSIONS

This review underscores a growing disconnect between the rapid evolution of AI technologies and the institutional maturity of tools suitable for scalable clinical integration. Despite notable technological advances, the clinical adoption of AI in telepsychiatry remains limited. The analysis identifies persistent methodological gaps and systemic barriers that demand coordinated efforts across research, technical, and regulatory communities. It also outlines key directions for future empirical studies and interdisciplinary development of implementation standards.

Perceived Impact of COVID-19 in an Underserved Community: A Natural Language Processing Approach

AIM

To utilise natural language processing (NLP) to analyse interviews about the impact of COVID-19 in underserved communities and to compare it to traditional thematic analysis in a small subset of interviews.

DESIGN

NLP and thematic analysis were used together to comprehensively examine the interview data.

METHODS

Fifty transcribed interviews with purposively sampled adults living in underserved communities in the United States, conducted from June 2021 to May 2022, were analysed to explore the impact of the COVID-19 pandemic on social activities, mental and emotional stress and physical and spiritual well-being. NLP includes several stages: data extraction, preprocessing, processing using word embeddings and topic modelling and visualisation. This was compared to thematic analysis in a random sample of 10 interviews.

RESULTS

Six themes emerged from thematic analysis: The New Normal, Juxtaposition of Emotions, Ripple Effects on Health, Brutal yet Elusive Reality, Evolving Connections and Journey of Spirituality and Self-Realisation. With NLP, four clusters of similar context words for each approach were analysed visually and numerically. The frequency-based word embedding approach was most interpretable and well aligned with the thematic analysis.

CONCLUSION

The NLP results complemented the thematic analysis and offered new insights regarding the passage of time, the interconnectedness of impacts and the semantic connections among words. This research highlights the interdependence of pandemic impacts, simultaneously positive and negative effects and deeply individual COVID-19 experiences in underserved communities.

IMPLICATIONS

The iterative integration of NLP and thematic analysis was efficient and effective, facilitating the analysis of many transcripts and expanding nursing research methodology.

IMPACT

While thematic analysis provided richer, more detailed themes, NLP captured new elements and combinations of words, making it a promising tool in qualitative analysis.

REPORTING METHOD

Not applicable.

PATIENT OR PUBLIC CONTRIBUTION

No patient or public contribution.

Detecting ADHD through natural language processing and stylometric analysis of adolescent narratives

Introduction

Attention-Deficit/Hyperactivity Disorder (ADHD) significantly affects adolescents' everyday lives, particularly in emotion regulation and interpersonal relationships. Despite its high prevalence, ADHD remains underdiagnosed, highlighting the need for improved diagnostic tools. This study explores, for the first time, the potential of Natural Language Processing (NLP) and stylometry to identify linguistic markers within Self-Defining Memories (SDMs) of adolescents with ADHD and to evaluate their utility in detecting the disorder. A further novel aspect of this research is the use of SDMs as a linguistic dataset, which reveals meaningful patterns while engaging psychological processes related to identity and memory.

Method

Our objectives were to: (1) characterize linguistic features of SDMs in ADHD and control groups; (2) assess the predictive power of stylometry in classifying participants' narratives as belonging to either the ADHD or control group; and (3) conduct a qualitative analysis of key linguistic markers of each group. Sixty-six adolescents (25 diagnosed with ADHD and 41 typically developing peers) recounted SDMs in a semi-structured format; these narratives were transcribed for analysis. Stylometric features were extracted and used to train a Support Vector Machine (SVM) classifier to distinguish between narratives from the ADHD and control groups. Linguistic metrics such as wordcount, lexical diversity, lexical density, and cohesion were computed and analyzed. A qualitative analysis was also applied to examine stylistic patterns in the narratives.

Results

Adolescents with ADHD produced narratives that were shorter, less lexically diverse, and less cohesive. Stylometric analysis using an SVM classifier distinguished between ADHD and control groups with up to 100% precision. Distinct linguistic markers were identified, potentially reflecting difficulties in emotion regulation.

Discussion

These findings suggest that NLP and stylometry can enhance ADHD diagnostics by providing objective linguistic markers, thereby improving both its understanding and diagnostic procedures. Further research is needed to validate these methods in larger and more diverse populations.

Towards a latent space cartography of subjective experience in mental health

AIMS

The way that individuals subjectively experience the world greatly influences their own mental well-being. However, it remains a considerable challenge to precisely characterize the breadth and depth of such experiences. One persistent problem is the lack of objective tools for directly quantifying and comparing narrative reports of subjective experiences. Here, we develop a new approach to map and compare reports of experience using the latent space of artificial neural networks.

METHODS

Using a series of 31 prompts, including 30 images and one open-ended question, we quantified how the verbal reports provided by participants (n = 210, 50% female) deviate from one another and how these variations are linked to subjective experience and mental health.

RESULTS

We found that latent space embeddings of experience can accurately predict subjective judgments of valence and arousal in a series of emotional pictures. Furthermore, we show that narrative reports to ambiguous images can accurately predict transdiagnostic factors of mental health. While distortions in the latent space of artificial neural networks are notoriously difficult to interpret, we propose a new approach to synthesize visual stimuli with generative artificial intelligence that can be used to explore semantic distortions in reported experiences.

CONCLUSIONS

In sum, latent space cartography could offer a promising avenue for objectively quantifying distortions of subjective experience in mental health and could ultimately help identify new therapeutic targets for clinical interventions.

Data transformation of unstructured electroencephalography reports by natural language processing: improving data usability for large-scale epilepsy studies

Introduction

Electroencephalography (EEG) is a popular technique that provides neurologists with electrographic insights and clinical interpretations. However, these insights are predominantly presented in unstructured textual formats, which complicates data extraction and analysis. In this study, we introduce a hierarchical algorithm aimed at transforming unstructured EEG reports from pediatric patients diagnosed with epilepsy into structured data using natural language processing (NLP) techniques.

Methods

The proposed algorithm consists of two distinct phases: a deep learning-based text classification followed by a series of rule-based keyword extraction procedures. First, we categorized the EEG reports into two primary groups: normal and abnormal. Thereafter, we systematically identified the key indicators of cerebral dysfunction or seizures, distinguishing between focal and generalized seizures, as well as identifying the epileptiform discharges and their specific anatomical locations. For this study, we retrospectively analyzed a dataset comprising 17,172 EEG reports from 3,423 pediatric patients. Among them, we selected 6,173 normal and 6,173 abnormal reports confirmed by neurologists for algorithm development.

Results

The developed algorithm successfully classified EEG reports into 1,000 normal and 1,000 abnormal reports, and effectively identified the presence of cerebral dysfunction or seizures within these reports. Furthermore, our findings revealed that the algorithm translated abnormal reports into structured tabular data with an accuracy surpassing 98.5% when determining the type of seizures (focal or generalized). Additionally, the accuracy for detecting epileptiform discharges and their respective locations exceeded 88.5%. These outcomes were validated through both internal and external assessments involving 800 reports from two different medical institutions.

Discussion

Our primary focus was to convert EEG reports into structured datasets, diverging from the traditional methods of formulating clinical notes or discharge summaries. We developed a hierarchical and streamlined approach leveraging keyword selections guided by neurologists, which contributed to the exceptional performance of our algorithm. Overall, this methodology enhances data accessibility as well as improves the potential for further research and clinical applications in the field of pediatric epilepsy management.

Machine learning tools match physician accuracy in multilingual text annotation

In the medical field, text annotation involves categorizing clinical and biomedical texts with specific medical categories, enhancing the organization and interpretation of large volumes of unstructured data. This process is crucial for developing tools such as speech recognition systems, which help medical professionals reduce their paperwork. It addresses a significant cause of burnout reported by up to 60% of medical staff. However, annotating medical texts in languages other than English poses unique challenges and necessitates using advanced models. In our research, conducted in collaboration with Gdańsk University of Technology and the Medical University of Gdańsk, we explore strategies to tackle these challenges. We evaluated the performance of various tools and models in recognizing medical terms within a comprehensive vocabulary, comparing these tools' outcomes with annotations made by medical experts. Our study specifically examined categories such as 'Drugs', 'Diseases and Symptoms', 'Procedures', and 'Other Medical Terms', contrasting human expert annotations with the performance of popular multilingual chatbots and natural language processing (NLP) tools on translated texts. The conclusion drawn from our statistical analysis reveals that no significant differences were detected between the groups we examined. This suggests that the tools and models we tested are, on average, similarly effective-or ineffective-at recognizing medical terms as categorized by our specific criteria. Our findings highlight the challenges in bridging the gap between human and machine accuracy in medical text annotation, especially in non-English contexts, and emphasize the need for further refinement of these technologies.

Co-producing a safe mobility and falls informatics platform to drive meaningful quality improvement in the hospital setting: a mixed-methods protocol for the insightFall study

INTRODUCTION

Manual investigation of falls incidents for quality improvement is time-consuming for clinical staff. Routine care delivery generates a large volume of relevant data in disparate systems, yet these data are seldom integrated and transformed into real-time, actionable insights for frontline staff. This protocol describes the co-design and testing of a safe mobility and falls informatics platform for automated, real-time insights to support the learning response to inpatient falls.

METHODS

Underpinned by the learning health system model and human-centred design principles, this mixed-methods study will involve (1) collaboration between healthcare professionals, patients, data scientists and researchers to co-design a safe mobility and falls informatics platform; (2) co-production of natural language processing pipelines and integration with a user interface for automated, near-real-time insights and (3) platform usability testing. Platform features (data taxonomy and insights display) will be co-designed during workshops with lay partners and clinical staff. The data to be included in the informatics platform will be curated from electronic health records and incident reports within an existing secure data environment, with appropriate data access approvals and controls. Exploratory analysis of a preliminary static dataset will examine the variety (structured/unstructured), veracity (accuracy/completeness) and value (clinical utility) of the data. Based on these initial insights and further consultation with lay partners and clinical staff, a final data extraction template will be agreed. Natural language processing pipelines will be co-produced, clinically validated and integrated with QlikView. Prototype testing will be underpinned by the Technology Acceptance Model, comprising a validated survey and think-aloud interviews to inform platform optimisation.

ETHICS AND DISSEMINATION

This study protocol was approved by the National Institute for Health Research Imperial Biomedical Research Centre Data Access and Prioritisation Committee (Database: iCARE-Research Data Environment; REC reference: 21/SW/0120). Our dissemination plan includes presenting our findings to the National Falls Prevention Coordination Group, publication in peer-reviewed journals, conference presentations and sharing findings with patient groups most affected by falls in hospital.

Natural language processing to identify suicidal ideation and anhedonia in major depressive disorder

BACKGROUND

Anhedonia and suicidal ideation are symptoms of major depressive disorder (MDD) that are not regularly captured in structured scales but may be captured in unstructured clinical notes. Natural language processing (NLP) techniques may be used to extract longitudinal data on suicidal behaviors and anhedonia within unstructured clinical notes. This study assessed the accuracy of using NLP techniques on electronic health records (EHRs) to identify these symptoms among patients with MDD.

METHODS

EHR-derived, de-identified data were used from the NeuroBlu Database (version 23R1), a longitudinal behavioral health real-world database. Mental health clinicians annotated instances of anhedonia and suicidal symptoms in clinical notes creating a ground truth. Interrater reliability (IRR) was calculated using Krippendorff's alpha. A novel transformer architecture-based NLP model was trained on clinical notes to recognize linguistic patterns and contextual cues. Each sentence was categorized into one of four labels: (1) anhedonia; (2) suicidal ideation without intent or plan; (3) suicidal ideation with intent or plan; (4) absence of suicidal ideation or anhedonia. The model was assessed using positive predictive values (PPV), negative predictive values, sensitivity, specificity, F1-score, and AUROC.

RESULTS

The model was trained, tested, and validated on 2,198, 1,247, and 1,016 distinct clinical notes, respectively. IRR was 0.80. For anhedonia, suicidal ideation with intent or plan, and suicidal ideation without intent or plan the model achieved a PPV of 0.98, 0.93, and 0.87, an F1-score of 0.98, 0.91, and 0.89 during training and a PPV of 0.99, 0.95, and 0.87 and F1-score of 0.99, 0.95, and 0.89 during validation.

CONCLUSIONS

NLP techniques can leverage contextual information in EHRs to identify anhedonia and suicidal symptoms in patients with MDD. Integrating structured and unstructured data offers a comprehensive view of MDD's trajectory, helping healthcare providers deliver timely, effective interventions. Addressing current limitations will further enhance NLP models, enabling more accurate extraction of critical clinical features and supporting personalized, proactive mental health care.

Medical Information Extraction With NLP-Powered QABots: A Real-World Scenario

The advent of computerized medical recording systems in healthcare facilities has made data retrieval tasks easier, compared to manual recording. Nevertheless, the potential of the information contained within medical records remains largely untapped, mostly due to the time and effort required to extract data from unstructured documents. Natural Language Processing (NLP) represents a promising solution to this challenge, as it enables the use of automated text-mining tools for clinical practitioners. In this work, we present the architecture of the Virtual Dementia Institute (IVD), a consortium of sixteen Italian hospitals, using the NLP Extraction and Management Tool (NEMT), a (semi-) automated end-to-end pipeline that extracts relevant information from clinical documents and stores it in a centralized REDCap database. After defining a common Case Report Form (CRF) across the IVD hospitals, we implemented NEMT, the core of which is a Question Answering Bot (QABot) based on a modern NLP model. This QABot is fine-tuned on thousands of examples from IVD centers. Detailed descriptions of the process to define a common minimum dataset, Inter-Annotator Agreement calculated on clinical documents, and NEMT results are provided. The best QABot performance show an Exact Match score (EM) of 78.1%, a F1-score of 84.7%, a Lenient Accuracy (LAcc) of 0.834, and a Mean Reciprocal Rank (MRR) of 0.810. EM and F1 scores outperform the same metrics obtained with ChatGPTv3.5 (68.9% and 52.5%, respectively). With NEMT the IVD has been able to populate a database that will contain data from thousands of Italian patients, all screened with the same procedure. NEMT represents an efficient tool that paves the way for medical information extraction and exploitation for new research studies.

Using a natural language processing toolkit to classify electronic health records by psychiatric diagnosis

Objective: We analyzed a natural language processing (NLP) toolkit's ability to classify unstructured EHR data by psychiatric diagnosis. Expertise can be a barrier to using NLP. We employed an NLP toolkit (CLARK) created to support studies led by investigators with a range of informatics knowledge. Methods: The EHR of 652 patients were manually reviewed to establish Depression and Substance Use Disorder (SUD) labeled datasets, which were split into training and evaluation datasets. We used CLARK to train depression and SUD classification models using training datasets; model performance was analyzed against evaluation datasets. Results: The depression model accurately classified 69% of records (sensitivity = 0.68, specificity = 0.70, F1 = 0.68). The SUD model accurately classified 84% of records (sensitivity = 0.56, specificity = 0.92, F1 = 0.57). Conclusion: The depression model performed a more balanced job, while the SUD model's high specificity was paired with a low sensitivity. NLP applications may be especially helpful when combined with a confidence threshold for manual review.

The Role of Virtual Reality and Artificial Intelligence in Cognitive Pain Therapy: A Narrative Review

PURPOSE OF REVIEW

This review investigates the roles of artificial intelligence (AI) and virtual reality (VR) in enhancing cognitive pain therapy for chronic pain management. The work assesses current research, outlines benefits and limitations and examines their potential integration into existing pain management methods.

RECENT FINDINGS

Advances in VR have shown promise in chronic pain management through immersive cognitive therapy exercises, with evidence supporting VR's effectiveness in symptom reduction. AI's personalization of treatment plans and its support for mental health through AI-driven avatars are emerging trends. The integration of AI in hybrid programs indicates a future with real-time adaptive technology tailored to individual needs in chronic pain management. Incorporating AI and VR into chronic pain cognitive therapy represents a promising approach to enhance management by leveraging VR's immersive experiences and AI's personalized tactics, aiming to improve patient engagement and outcomes. Nonetheless, further empirical studies are needed to standardized methodologies, compare these technologies to traditional therapies and fully realize their clinical potential.

Natural Language Processing Applied to Spontaneous Recall of Famous Faces Reveals Memory Dysfunction in Temporal Lobe Epilepsy Patients

Objective and Background

Epilepsy patients rank memory problems as their most significant cognitive comorbidity. Current clinical assessments are laborious to administer and score and may not always detect subtle memory decline. The Famous Faces Task (FF) has robustly demonstrated that left temporal lobe epilepsy (LTLE) patients remember fewer names and biographical details compared to right TLE (RTLE) patients and healthy controls (HCs). We adapted the FF task to capture subjects' entire spontaneous spoken recall, then scored responses using manual and natural language processing (NLP) methods. We expected to replicate previous group level differences using spontaneous speech and semi-automated analysis.

Methods

Seventy-three (N=73) adults (28 LTLE, 18 RTLE, and 27 HCs) were included in a case-control prospective study design. Twenty FF in politics, sports, and entertainment (active 2008-2017) were shown to subjects, who were asked if they could recognize and spontaneously recall as much biographical detail as possible. We created human-generated and automatically-generated keyword dictionaries for each celebrity, based on a randomly selected training set of half of the HC transcripts. To control for speech output, we measured the speech duration, total word count and content word count for the FF task and a Cookie Theft Control Task (CTT), in which subjects were merely asked to describe a visual scene. Subjects' responses to FF and CTT tasks were recorded, transcribed, and analyzed in a blinded manner with a combination of manual and automated NLP approaches.

Results

Famous face recognition accuracy was similar between groups. LTLE patients recalled fewer biographical details compared to HCs and RTLEs using both the gold-standard human-generated dictionary (24%±12% vs. 31%±12% and 30%±12%, p=0.007) and the automated dictionary (24%±12% vs. 31%±12% and 32%±13%, p=0.007). There were no group level differences in speech duration, total word count, or content word count for either the FF and CTT to explain difference in recall performance. There was a positive, statistically significant relationship between MOCA score and FF recall performance as scored by the human-generated (ρ= .327, p= .029) and automatically-generated dictionaries (ρ= .422, p= .004) for TLE subjects, but not HCs, an effect that was driven by LTLE subjects.

Discussion

LTLE patients remember fewer details of famous people than HCs or RTLE patients, as discovered by NLP analysis of spontaneous recall. Decreased biographical memory was not due to decreased speech output and correlated with lower MOCA scores. NLP analysis of spontaneous recall can detect memory dysfunction in clinical populations in a semi-automated, objective, and sensitive manner.

Natural Language Processing and Schizophrenia: A Scoping Review of Uses and Challenges

(1) Background: Approximately 1% of the global population is affected by schizophrenia, a disorder marked by cognitive deficits, delusions, hallucinations, and language issues. It is associated with genetic, neurological, and environmental factors, and linked to dopaminergic hyperactivity and neurotransmitter imbalances. Recent research reveals that patients exhibit significant language impairments, such as reduced verbal output and fluency. Advances in machine learning and natural language processing show potential for early diagnosis and personalized treatments, but additional research is required for the practical application and interpretation of such technology. The objective of this study is to explore the applications of natural language processing in patients diagnosed with schizophrenia. (2) Methods: A scoping review was conducted across multiple electronic databases, including Medline, PubMed, Embase, and PsycInfo. The search strategy utilized a combination of text words and subject headings, focusing on schizophrenia and natural language processing. Systematically extracted information included authors, population, primary uses of the natural language processing algorithms, main outcomes, and limitations. The quality of the identified studies was assessed. (3) Results: A total of 516 eligible articles were identified, from which 478 studies were excluded based on the first analysis of titles and abstracts. Of the remaining 38 studies, 18 were selected as part of this scoping review. The following six main uses of natural language processing were identified: diagnostic and predictive modeling, followed by specific linguistic phenomena, speech and communication analysis, social media and online content analysis, clinical and cognitive assessment, and linguistic feature analysis. (4) Conclusions: This review highlights the main uses of natural language processing in the field of schizophrenia and the need for more studies to validate the effectiveness of natural language processing in diagnosing and treating schizophrenia.

Characterizing Myasthenia Gravis Symptoms, Exacerbations, and Crises From Neurologist's Clinical Notes Using Natural Language Processing

Background Myasthenia gravis (MG) is a rare, autoantibody neuromuscular disorder characterized by fatigable weakness. Real-world evidence based on administrative and structured datasets regarding MG may miss important details related to the clinical encounter. Examination of free-text clinical progress notes has the potential to illuminate aspects of MG care. Objective The primary objective was to examine and characterize neurologist progress notes in the care of individuals with MG regarding the prevalence of documentation of clinical subtypes, antibody status, symptomatology, and MG deteriorations, including exacerbations and crises. The secondary objectives were to categorize MG deteriorations into practical, objective states as well as examine potential sources of clinical inertia in MG care. Methods We performed a retrospective, cross-sectional analysis of de-identified neurologist clinical notes from 2017 to 2022. A qualitative analysis of physician descriptions of MG deteriorations and a discussion of risks in MG care (risk for adverse effects, risk for clinical decompensation, etc.) was performed. Results Of the 3,085 individuals with MG, clinical subtypes and antibody status identified included gMG (n = 400; 13.0%), ocular MG (n = 253; 8.2%), MG unspecified (2,432; 78.8%), seropositivity for acetylcholine receptor antibody (n = 441; 14.3%), and MuSK antibody (n = 29; 0.9%). The most common gMG manifestations were dysphagia (n = 712; 23.0%), dyspnea (n = 626; 20.3%), and dysarthria (n = 514; 16.7%). In MG crisis patients, documentation of difficulties with MG standard therapies was common (n = 62; 45.2%). The qualitative analysis of MG deterioration types includes symptom fluctuation, symptom worsening with treatment intensification, MG deterioration with rescue therapy, and MG crisis. Qualitative analysis of MG-related risks included the toxicity of new therapies and concern for worsening MG because of changing therapies. Conclusions This study of neurologist progress notes demonstrates the potential for real-world evidence generation in the care of individuals with MG. MG patients suffer fluctuating symptomatology and a spectrum of clinical deteriorations. Adverse effects of MG therapies are common, highlighting the need for effective, less toxic treatments.

GPT-4 Performance for Neurologic Localization

Background and Objectives

In health care, large language models such as Generative Pretrained Transformers (GPTs), trained on extensive text datasets, have potential applications in reducing health care disparities across regions and populations. Previous software developed for lesion localization has been limited in scope. This study aims to evaluate the capability of GPT-4 for lesion localization based on clinical presentation.

Methods

GPT-4 was prompted using history and neurologic physical examination (H&P) from published cases of acute stroke followed by questions for clinical reasoning with answering for "single or multiple lesions," "side," and "brain region" using Zero-Shot Chain-of-Thought and Text Classification prompting. GPT-4 output on 3 separate trials for each of 46 cases was compared with imaging-based localization.

Results

GPT-4 successfully processed raw text from H&P to generate accurate neuroanatomical localization and detailed clinical reasoning. Performance metrics across trial-based analysis for specificity, sensitivity, precision, and F1-score were 0.87, 0.74, 0.75, and 0.74, respectively, for side; 0.94, 0.85, 0.84, and 0.85, respectively, for brain region. Class labels within the brain region were similarly high for all regions except the cerebellum and were also similar when considering all 3 trials to examine metrics by case. Errors were due to extrinsic causes-inadequate information in the published cases, and intrinsic causes-failures of logic or inadequate knowledge base.

Discussion

This study reveals capabilities of GPT-4 in the localization of acute stroke lesions, showing a potential future role as a clinical tool in neurology.

A Health Care Clinical Data Platform for Rapid Deployment of Artificial Intelligence and Machine Learning Algorithms for Cancer Care and Oncology Clinical Trials

The xCures platform aggregates, organizes, structures, and normalizes clinical EMR data across care sites, utilizing advanced technologies for near real-time access. The platform generates data in a format to support clinical care, accelerate research, and promote artificial intelligence/ machine learning algorithm development, highlighted by a clinical decision support algorithm for precision oncology.

Large Language Models in Neurology Research and Future Practice

Recent advancements in generative artificial intelligence, particularly using large language models (LLMs), are gaining increased public attention. We provide a perspective on the potential of LLMs to analyze enormous amounts of data from medical records and gain insights on specific topics in neurology. In addition, we explore use cases for LLMs, such as early diagnosis, supporting patient and caregivers, and acting as an assistant for clinicians. We point to the potential ethical and technical challenges raised by LLMs, such as concerns about privacy and data security, potential biases in the data for model training, and the need for careful validation of results. Researchers must consider these challenges and take steps to address them to ensure that their work is conducted in a safe and responsible manner. Despite these challenges, LLMs offer promising opportunities for improving care and treatment of various neurologic disorders.

Identifying encephalopathy in patients admitted to an intensive care unit: Going beyond structured information using natural language processing

Background

Encephalopathy is a severe co-morbid condition in critically ill patients that includes different clinical constellation of neurological symptoms. However, even for the most recognised form, delirium, this medical condition is rarely recorded in structured fields of electronic health records precluding large and unbiased retrospective studies. We aimed to identify patients with encephalopathy using a machine learning-based approach over clinical notes in electronic health records.

Methods

We used a list of ICD-9 codes and clinical concepts related to encephalopathy to define a cohort of patients from the MIMIC-III dataset. Clinical notes were annotated with MedCAT and vectorized with a bag-of-word approach or word embedding using clinical concepts normalised to standard nomenclatures as features. Machine learning algorithms (support vector machines and random forest) trained with clinical notes from patients who had a diagnosis of encephalopathy (defined by ICD-9 codes) were used to classify patients with clinical concepts related to encephalopathy in their clinical notes but without any ICD-9 relevant code. A random selection of 50 patients were reviewed by a clinical expert for model validation.

Results

Among 46,520 different patients, 7.5% had encephalopathy related ICD-9 codes in all their admissions (group 1, definite encephalopathy), 45% clinical concepts related to encephalopathy only in their clinical notes (group 2, possible encephalopathy) and 38% did not have encephalopathy related concepts neither in structured nor in clinical notes (group 3, non-encephalopathy). Length of stay, mortality rate or number of co-morbid conditions were higher in groups 1 and 2 compared to group 3. The best model to classify patients from group 2 as patients with encephalopathy (SVM using embeddings) had F1 of 85% and predicted 31% patients from group 2 as having encephalopathy with a probability >90%. Validation on new cases found a precision ranging from 92% to 98% depending on the criteria considered.

Conclusions

Natural language processing techniques can leverage relevant clinical information that might help to identify patients with under-recognised clinical disorders such as encephalopathy. In the MIMIC dataset, this approach identifies with high probability thousands of patients that did not have a formal diagnosis in the structured information of the EHR.