How ChatGPT works: a mini review.

OBJECTIVE: This paper offers a mini-review of OpenAI's language model, ChatGPT, detailing its mechanisms, applications in healthcare, and comparisons with other large language models (LLMs). METHODS: The underlying technology of ChatGPT is outlined, focusing on its neural network architecture, training process, and the role of key elements such as input embedding, encoder, decoder, attention mechanism, and output projection. The advancements in GPT-4, including its capacity for internet connection and the integration of plugins for enhanced functionality are discussed. RESULTS: ChatGPT can generate creative, coherent, and contextually relevant sentences, making it a valuable tool in healthcare for patient engagement, medical education, and clinical decision support. Yet, like other LLMs, it has limitations, including a lack of common sense knowledge, a propensity for hallucination of facts, a restricted context window, and potential privacy concerns. CONCLUSION: Despite the limitations, LLMs like ChatGPT offer transformative possibilities for healthcare. With ongoing research in model interpretability, common-sense reasoning, and handling of longer context windows, their potential is vast. It is crucial for healthcare professionals to remain informed about these technologies and consider their ethical integration into practice.

Accuracy of ChatGPT-3.5 and -4 in providing scientific references in otolaryngology-head and neck surgery.

INTRODUCTION: Chatbot generative pre-trained transformer (ChatGPT) is a new artificial intelligence-powered language model of chatbot able to help otolaryngologists in practice and research. We investigated the accuracy of ChatGPT-3.5 and -4 in the referencing of manuscripts published in otolaryngology. METHODS: ChatGPT-3.5 and ChatGPT-4 were interrogated for providing references of the top-30 most cited papers in otolaryngology in the past 40 years including clinical guidelines and key studies that changed the practice. The responses were regenerated three times to assess the accuracy and stability of ChatGPT. ChatGPT-3.5 and ChatGPT-4 were compared for accuracy of reference and potential mistakes. RESULTS: The accuracy of ChatGPT-3.5 and ChatGPT-4.0 ranged from 47% to 60%, and 73% to 87%, respectively (p < 0.005). ChatGPT-3.5 provided 19 inaccurate references and invented 2 references throughout the regenerated questions. ChatGPT-4.0 provided 13 inaccurate references, while it proposed only one invented reference. The stability of responses throughout regenerated answers was mild (k = 0.238) and moderate (k = 0.408) for ChatGPT-3.5 and 4.0, respectively. CONCLUSIONS: ChatGPT-4.0 reported higher accuracy than the free-access version (3.5). False references were detected in both 3.5 and 4.0 versions. Practitioners need to be careful regarding the use of ChatGPT in the reach of some key reference when writing a report.

Machine learning for risk stratification of thyroid cancer patients: a 15-year cohort study.

PURPOSE: The objective of this study was to train machine learning models for predicting the likelihood of recurrence in patients diagnosed with well-differentiated thyroid cancer. While thyroid cancer mortality remains low, the risk of recurrence is a significant concern. Identifying individual patient recurrence risk is crucial for guiding subsequent management and follow-ups. METHODS: In this prospective study, a cohort of 383 patients was observed for a minimum duration of 10 years within a 15-year timeframe. Thirteen clinicopathologic features were assessed to predict recurrence potential. Classic (K-nearest neighbors, support vector machines (SVM), tree-based models) and artificial neural networks (ANN) were trained on three distinct combinations of features: a data set with all features excluding American Thyroid Association (ATA) risk score (12 features), another with ATA risk alone, and a third with all features combined (13 features). 283 patients were allocated for the training process, and 100 patients were reserved for the validation of stage. RESULTS: The patients' mean age was 40.87 ± 15.13 years, with a majority being female (81%). When using the full data set for training, the models showed the following sensitivity, specificity and AUC, respectively: SVM (99.33%, 97.14%, 99.71), K-nearest neighbors (83%, 97.14%, 98.44), Decision Tree (87%, 100%, 99.35), Random Forest (99.66%, 94.28%, 99.38), ANN (96.6%, 95.71%, 99.64). Eliminating ATA risk data increased models specificity but decreased sensitivity. Conversely, training exclusively on ATA risk data had the opposite effect. CONCLUSIONS: Machine learning models, including classical and neural networks, efficiently stratify the risk of recurrence in patients with well-differentiated thyroid cancer. This can aid in tailoring treatment intensity and determining appropriate follow-up intervals.

Online software Boneureka assessing bone age based on metacarpal length in healthy children: proof-of-concept study.

BACKGROUND: Bone age in children is mainly assessed using the Greulich and Pyle (GP) atlas, a validated method with limited interobserver accuracy. While automated methods increase interobserver accuracy, they represent considerable costs and technical requirements. OBJECTIVE: A proof-of-concept study to create and evaluate an online software program, Boneureka©, based on linear metacarpal length measurements, to assess bone age in healthy children. MATERIALS AND METHODS: The study retrospectively included 434 consecutive children (215 girls) who underwent a left-hand radiograph to rule out trauma between March 2008 and December 2017. Two reviewers measured the second to fourth metacarpal lengths on each radiograph and the distance between the centre of the epiphyses of the second and fifth metacarpals. A single reviewer estimated the bone age using the GP atlas. The automated software assessed the bone age for all radiographs. A mathematical model was developed based on linear regressions to provide the mean bone age and standard deviation based on the estimates. Pearson and intraclass correlation coefficient (ICC) were used to evaluate the correlation and agreement between the estimated bone ages using Boneureka©, the GP atlas and BoneXpert® compared to chronological age. RESULTS: The measure that showed the highest correlation (r2=0.877 for girls and r2=0.834 for boys; P<.001) and the highest ICC (ICC=0.937 for girls and ICC=0.926 for boys; P<0.001) with chronological age was length of the second metacarpal. The GP atlas and the automated software evaluation had excellent ICC with chronological age (ICC>0.95 for both methods and sexes). Using this data, we created an online software program based on the second metacarpal length to obtain bone age estimates, means and standard deviations. CONCLUSION: The newly created online software Boneureka,© based on the second metacarpal length, is a reliable and user-friendly tool to assess bone age in healthy children. Further studies on a larger population should be performed to validate the developed reference values.

Artificial Intelligence in Psychiatry.

Artificial Intelligence (AI) has emerged as a powerful tool in various fields, including psychiatry. This paper explores the potential of AI in the diagnosis, treatment, and understanding of psychiatric conditions. We delve into the role of AI in psychiatry, discussing its applications, challenges, and future directions. We explore how AI techniques such as classification, hypothesis generation, and prediction are being used in psychiatry, with a specific focus on the detection and prediction of psychiatric conditions. We also discuss the ethical considerations and challenges in implementing AI in psychiatry and look towards the future of AI in this field. The paper highlights the potential of AI to enhance our understanding of psychiatric conditions, improve patient care, and drive innovation in psychiatric research. However, it also underscores the need for robust ethical frameworks and stringent data protection measures to ensure the responsible and effective use of AI in psychiatry.

Neuropsychiatry and Neuropsychopharmacology of Dementia: a Guide to Evaluation, Diagnosis, and Management.

In the present study, I provide an examination of the neuropsychiatric approach to patients with various types of dementia, including Alzheimer's disease, Parkinson's disease dementia, Lewy body dementia, vascular dementia, frontotemporal dementia, and more. With a focus on the intersection of psychiatry and neurology, this paper underscores the importance of comprehensive neuropsychiatric evaluation, rigorous diagnosis, and evidence-based management. The paper delineates the neuropsychiatric manifestations specific to each type of dementia and explores both non-pharmacological and pharmacological management strategies, aiming to equip psychiatrists with the latest evidence-based approaches. Case studies are included to demonstrate real-world clinical scenarios and to provide insights into the practical application of the theories discussed. Additionally, this guide addresses current challenges in the neuropsychiatric approach to dementia and highlights potential solutions and future research directions. The primary objective of this guide is to enable psychiatrists to enhance the quality of life for individuals living with dementia by improving understanding, diagnosis, and management of the neuropsychiatric aspects of these conditions.

Narrating: from Evolutionary to Clinical Advantages.

Human beings constantly narrate reality. They narrate themselves, to themselves and to others. They narrate each other and narrate humanity. They narrate the world and nature. They narrate meaning, the meaning of life and things. This article aims to explore this phenomenon of "narrating". Through a narrative review, we will attempt to gather elements of reflection on narrative, considered here as the ability to narrate, it means to represent oneself, to put meaning. Firstly, we will focus on how cognition, interpretation, and culture allow Homo Sapiens to narrate reality to himself. Then, we will explore why they do it and discover the evolutionary advantages of language, of sharing experiences between individuals through the phenomenon of cumulative cultural evolution, and how narrative facilitates the species' access to these advantages. Finally, we will delve into the clinical implications of narrative, discussing therapeutic interviews, therapy, and psychopathology. Narratives and pre-linguistic mental representations appear to have driven the Homo genus to develop cognitive abilities that enable the development of language and the sophistication of narratives as a cultural medium. Through language, Homo sapiens share their narratives and develop a cumulative common culture. Each individual's culture seems to be constructed in dialectic with this shared culture through narratives. This dialectic gives rise to psychopathological processes while also producing therapeutic leverage. Understanding the mechanisms of co-construction of these narratives is essential in clinical research in mental health. Furthermore, placing narratives in the perspective of an essential evolutionary strategy in the Homo genus solidifies the significance of the narrative faculty in the biological functioning of Homo sapiens, and so the importance of narratives in mental health.

AI in Child Psychiatry: Exploring Future Tools for the Detection and Management of Mental Disorders in Children and Adolescents.

This article explores the potential of artificial intelligence (AI) in the field of child psychiatry, focusing on screening, detection, diagnosis, therapeutic tools, and research development. In this non-exhaustive review, we will examine the interest of AI applications in intervention contexts such as parent education tools, emotion regulation tools, and cognitive-behavioral therapy tools through interactive applications. The network approach, a branch of machine learning, will also be considered. Some examples of machine learning applications in child psychiatry will be presented. Finally, we address the ethical question of the role of these applications, analyzing whether they represent a real beneficial tool or a potential danger.

Empathy and Psychopathy Traits in University Students in Belgium: a Network Approach.

Empathy is the complex prosocial cognitive capacity to recognize and react to the emotions of others. An empathic attitude from medical doctors is essential to build a good relationship with patients. In engineering education there is an hypervalorization of technical skills in disadvantage of these affective elements. Psychopathy is the lack of considerations toward others. These two important personality traits shape social interactions. In this study we analyzed, through the network theory, these characteristics in a young population of medical and engineering university students in Belgium. The aim of this study was therefore to estimate the individual network structure of these traits in both groups, as well as estimate whether there is a fundamental difference in the way that these traits connect in these two populations. Medical and engineering students completed online three self-report questionnaires about empathy and psychopathy traits. There were 178 responders without exclusions due outliers. No significant differences were found in psychopathic traits between the two groups. There was a statistically significant difference in empathic concerns, the medical students being more empathic than their peers in engineering. Psychopathic traits did not vary significantly between the two groups. This study provided insights into the differences in empathic and psychopathic traits among those students. Future research should explore the factors that contribute to these differences and investigate the potential impact of targeted interventions or curricular modifications in cultivating empathy and minimizing antisocial behaviors in both fields.

Somatic Management of Psychiatric Patients in the Emergency Department.

This retrospective study aims to assess a potential difference in the management of patients with a psychiatric history in somatic emergencies. Indeed, the psychiatric population has higher mortality and morbidity rates than the general population. The negative stigmatization of patients with mental health disorders remains one of the factors to consider when studying this morbidity and mortality. In this context, adult patients diagnosed with myocardial infarction, pulmonary embolism, stroke, acute cholecystitis or appendicitis in the emergency department of the Brugmann University Hospital Center during the year 2021 were selected. The presence or absence of a history psychiatric was then recorded for each patient. Different key intervention times, the total length of stay and the occurrence of complications were also studied for 459 patients, 74 of which had a history psychiatric. A significant difference in the time preceding the prescription of the first complementary examination for patients with a psychiatric history was thus highlighted. No other differences in care were demonstrated within the limits of this sample. This difference could be associated with the phenomenon of diagnostic overshadowing. It is the fact of associating the somatic complaints of a patient with his psychiatric pathology. Another potential explanation, present in the literature, could be the discomfort felt by somaticians when dealing with psychiatric patients. Finally, the integration of the experience of psychiatric patients into the training of physicians and the question of the relevance of applying the triage system to psychiatric patients were raised as potential future studies.

On the Use of Bayesian Artificial Intelligence for Hypothesis Generation in Psychiatry.

In this study, I introduce the use of Bayesian Artificial Intelligence, namely through the probabilistic and structure learning of Bayesian Network models, for hypothesis generation in psychiatry. Bayesian Networks are directed acyclic graphical models that allow researchers to account for complexity in multivariate data sets, as well as identify what is the likely causal direction in detected associations. This in turns leads to more effective designs for confirmatory studies in clinical settings, that go beyond association studies and can provide meaningful impact in clinical practice. As an example, I use three different data sets to highlight several frameworks for hypothesis generation. Bayesian Networks are useful models since the early stages of knowledge generation in psychiatry, and they can be easily adopted by most applied and clinical researchers for use in quantitative studies.

Investigation and Comparison of Networks of Psychopathic Traits in Psychiatric Inpatients and University Students.

The Levenson Self-Report Psychopathy Scale (LSRPS) is a psychometric tool composed of 26 items to assess psychopathic traits. This study aims to perform a network analysis of this scale in a large sample composed of 100 hospitalized psychiatric patients and 256 French-speaking Belgian university students in medicine and to compare the network structure. We estimated a regularized partial correlation network for the 26 items of the questionnaire. Node predictability is used to assess the connectivity of items. The network comparison test (NCT) and statistical inference on sum scores are conducted to compare networks from the inpatients and the university students. The networks composed of LSRPS are mostly connected positively, but some negative interconnections were observed in both inpatients and university students, and node connectivity varies. Although the scores from inpatients are substantially higher than those of university students, network analysis didn't show any statistical difference in the overall connectedness. Network analysis is a valuable tool for exploring psychopathic traits and offers new insight into how they interact. In the network estimation, we concluded that the two domains of psychopathy are interrelated. This interconnectivity was observed in both subject groups. We hypothesize that such interconnectivity was present because environmental and genetic factors are intricately intertwined in the appearance of primary and secondary psychopathy. Meanwhile, although inpatients may have higher scores of psychopathic traits, those traits' connectedness isn't different from that of the general population. This finding aligns well with the theory of hysteresis in network analysis, which states that the connections among components of mental disorders do not disappear or reappear over time, but their importance may vary. Further studies may replicate our findings using different sample groups.

The Borders of Mental Illness: Investigating Clusters of Diagnostic Criteria in Acute Presentations of Psychiatric Disorders.

The impact of diagnostic delay in mental health is significant in terms of neurocognitive impairment, comorbidities, prognostic and socio-economical cost. For this reason, diagnostical research in psychiatry and the classification methods are continuously questioned. The network theory of mental disorders aims at contributing to the improvement of psychiatric diagnosis and considers mental disorders as the results of complex sets of interactions among symptoms instead of being their common cause. In this study, we use network theory and its associate statistical methods, namely Gaussian Graphical Models, centrality, and cluster analysis, to estimate respectively the interactions among symptoms from different disorders, their relative importance, and how they overlap, in a sample of psychiatric inpatients. The community detection found nine clusters with their interactions. Many are closely related to DSM criteria but some of them share symptoms from both diagnostics. One central symptom of the construct is Insomnia. There was a significant difference in the sum scores for psychotic symptoms, but not for bipolar symptoms, across psychotic and bipolar patients. This study needs however to be replicated in a bigger sample of different patients. Computing Bayesian Network to assess causalities in the network and adding other variables (such as biomarkers or therapeutic responses) could contribute to a more personalized diagnostic. How symptoms connect to each other in a specific time frame would define a person phenotype. Network analysis allows for investigating connections, identifying which symptoms are relatively important to the self-determination of disorders as well as how network nodes predict each other and arise in communities. For instance, in psychotic and mood disorders, sleep related symptoms or altered speech features and the importance of their communities in the probable transfer of symptomatology.

Teen Dating Violence: A Mixed Methods Study in French-speaking Belgium.

BACKGROUND: "Teen dating violence" is a poorly studied issue in the literature. Several sociodemographic and psychiatric risk factors have been identified in the literature, and the consequences are numerous. The aim of this mixed study is to establish an initial inventory of the problem in French-speaking Belgium, to establish links with the socio-demographic and overall health characteristics of the population, and to issue recommendations for the future. SUBJECTS AND METHODS: A study by questionnaire was conducted among 103 young people aged between 14 and 18 living in French-speaking Belgium and speaking French in particular. Qualitative and quantitative (descriptive and inferential statistics) analyses were performed. RESULTS: 29.1% of respondents experienced violence. Significant links exist between the fact of having experienced violence and consumption as well as the presence of STDs. The combination of potential risk factors predicts a considerable part of the violence in the sample. A significant link exists between the fact of being subjected to violence and a high score on the scale of depression and violence suffered and committed. CONCLUSION: The prevalence of teen dating violence is relatively high in Belgium. A link between these and a large part of the potential consequences studied could be highlighted through the study.

Unique Variable Analysis of Redundancy in ADHD Items from the Conners Teacher Rating Scale - Revised: Short.

Attention deficit hyperactivity disorder (ADHD) is a neuropsychiatric disorder interfering with the normal development of the child. The disorder can be screened at school with the Conners Teacher Rating Scale Revised Short (CTRS-R:S). This scale goes beyond the disorder itself and covers a wider construct, that of abnormal child behavior. This can be understood as a complex system of mutually influencing entities. We analyzed a data set of 525 children in French-speaking primary schools from Belgium, and estimated a network structure, as well as to determine the local dependence of items through Unique Variable Analysis. A reduced network was computed including 15 non-locally dependent items. The structural consistency of the network was not affected by redundant items and was structurally sound. The reduction of the number of variables in network studies is important to improve the investigation of network structures as well as better interpret results from inference measures.

Appropriateness for SARS-CoV-2 vaccination for otolaryngologist and head and neck surgeons in case of pregnancy, breastfeeding, or childbearing potential: Yo-IFOS and CEORL-HNS joint clinical consensus statement.

PURPOSE: SARS-CoV-2 vaccines are a key step in fighting the pandemic. Nevertheless, their rapid development did not allow for testing among specific population subgroups such as pregnant and breastfeeding women, or elaborating specific guidelines for healthcare personnel working in high infection risk specialties, such as otolaryngology (ORL). This clinical consensus statement (CCS) aims to offer guidance for SARS-CoV-2 vaccination to this high-risk population based on the best evidence available. METHODS: A multidisciplinary international panel of 33 specialists judged statements through a two-round modified Delphi method survey. Statements were designed to encompass the following topics: risk of SARS-Cov-2 infection and use of protective equipment in ORL; SARS-Cov-2 infection and vaccines and respective risks for the mother/child dyad; and counseling for SARS-CoV-2 vaccination in pregnant, breastfeeding, or fertile healthcare workers (PBFHW). All ORL PBFHW were considered as the target audience. RESULTS: Of the 13 statements, 7 reached consensus or strong consensus, 2 reached no consensus, and 2 reached near-consensus. According to the statements with strong consensus otorhinolaryngologists-head and neck surgeons who are pregnant, breastfeeding, or with childbearing potential should have the opportunity to receive SARS-Cov-2 vaccination. Moreover, personal protective equipment (PPE) should still be used even after the vaccination. CONCLUSION: Until prospective evaluations on these topics are available, ORL-HNS must be considered a high infection risk specialty. While the use of PPE remains pivotal, ORL PBFHW should be allowed access to SARS-CoV-2 vaccination provided they receive up-to-date information.

A Machine Learning Approach to Relationships Among Alexithymia Components.

BACKGROUND: The aim of this paper is to explore the network structures of alexithymia components and compare results with relevant prior literature. SUBJECTS AND METHODS: In a large sample of university students, undirected and directed network structures of items from the Bermond Vorst Alexithymia Questionnaire form B are estimated with state-of-the-art network analysis and structure learning tools. Centrality estimates are used to address the topic of item redundancy and select relevant alexithymia components to study. RESULTS: Alexithymia components present positive as well as negative connections; poor fantasy and emotional insight are identified as central items in the network. CONCLUSIONS: The undirected network structure of alexithymia components reports new features with respect to prior literature, and the directed network structures offers new insight on the construct.

The Utilisation of Digital Applications for Measuring Patient Outcomes Following Bariatric Surgery: a Systematic Review and Meta-analysis of Comparative Studies.

In the context of escalating obesity rates, bariatric surgery holds a crucial role in managing severely obese patients. With a demonstrated effectiveness in weight loss and with the advent of ambulatory surgery, bariatric surgery allows for a streamlined care pathway, ideally suited for postoperative surveillance using digital health applications. The aim of this systematic review and meta-analysis is to evaluate the effect of eHealth-delivered health services or support for adults undergoing bariatric surgery. Five studies, encompassing 2210 patients, were analysed. The intervention group showed a 10% increase in total weight reduction and a 22% reduction in excess weight loss. ED visitation rates also trended towards reduction. Despite the absence of clear statistical superiority for DHA, the findings suggest potential benefits of DHA in postoperative monitoring.

Digestive Biomarkers of Laryngopharyngeal Reflux: A Preliminary Prospective Controlled Study.

OBJECTIVE: To investigate the digestive enzymes and biomarkers in the saliva of patients with laryngopharyngeal reflux (LPR) and asymptomatic individuals. STUDY DESIGN: Prospective controlled study. SETTING: Multicenter study. METHODS: Patients with LPR at the hypopharyngeal-esophageal impedance-pH monitoring (HEMII-pH) and asymptomatic individuals were consecutively recruited from January 2020 to April 2023 from 2 University Hospitals. The saliva of patients (off PPIs) and asymptomatic individuals was collected to measure pH, elastase, bile salts, cholesterol, gastric, and pancreatic lipases. Anxiety, symptoms, and findings were studied through perceived stress scale (PSS), reflux symptom score (RSS), and reflux sign assessment (RSA). RESULTS: Sixty-seven LPR patients and 57 asymptomatic individuals completed the evaluations. LPR patients reported higher PSS, RSS, and RSA than asymptomatic individuals. The mean saliva pH was more alkaline in LPR patients (7.23: 95% confidence interval [CI]: 7.08, 7.38) compared to controls (6.13; 95% CI: 5.95, 6.31; P = .001). The mean concentration of elastase was higher in patients (51.65 µg/mL; 95% CI: 44.47, 58.83 µg/mL) versus asymptomatic individuals (25.18 µg/mL; 95% CI: 21.64, 28.72 µg/mL; P = .001). The saliva cholesterol reported higher concentration in healthy individuals (3.43 mg/dL; 95% CI: 3.21, 3.65 mg/dL) compared to patients (1.16 mg/dL; 95% CI: 1.05, 1.27 mg/dL; P = .001). The saliva pH, and elastase concentration were significantly associated with the baseline RSS, while saliva cholesterol was negatively associated with the severity of RSS and RSA. CONCLUSION: Cholesterol, bile salts, and elastase are biomarkers of LPR and should be considered to develop future non-invasive saliva device for the detection of LPR.