Aim: The primary aim was to develop and test a telemedicine program for oral cancer screening by dentists in primary care. The secondary aim was to analyze the sensitivity of the provisional diagnosis compared to the definitive diagnosis.
Methods: A retrospective observational study that used telemedicine for oral cancer case detection was conducted in Cordoba, Argentina from 2018 to 2023, oral medicine specialists provided in-person training for dentists on the clinical recognition and early diagnosis of oral squamous cell carcinoma (OSCC) and oral potentially malignant disorders (OPMD), and telemedicine use for the early detection of oral cancer. The trained professionals conducted opportunistic screenings in their workplaces. When encountering a suspicious lesion on the oral mucosa, they collected relevant patient data and clinical photographs of the lesion, sharing these with the reference center. The specialized center was based at the Oral Medicine unit at the Facultad de Odontología, Universidad Nacional de Córdoba, Argentina. The specialists suggested radiographic examinations and/or pre-surgical laboratory tests and, if necessary, expedited referral to the specialized center for in-person assessment and definitive diagnosis.
Results: Cases with clinical suspicion of OSCC and OPMD were referred to the reference center. In all cases, the definitive diagnosis was obtained within less than 1 month. Eleven out of 12 cases of OSCC were diagnosed within 2 weeks, with only 1 case diagnosed at 1 month due to some patient delay. The concordance between the clinical suspicion at the time of teleconsultation and the definitive diagnosis of OSCC by the specialists was absolute (Kappa test, coefficient 1), with a sensitivity and specificity of 100%.
Conclusions: Integrating telemedicine with other preventive strategies and timely referral to oral medicine specialists could potentially decrease diagnostic delays in OSCC and OPMD.
Aim: Physical activity of nursing home residents can be assessed with tools such as questionnaires and standardized fitness tests. For residents with dementia, however, those tools can be cognitively challenging and difficult to administer. Consumer wearables could potentially aid as an affordable tool for ubiquitous assessment.
Methods: In this pilot study with 16 participants, we explored how measurements with an off-the-shelf wearable relate to structured observations of physical activity. We collected both processed and raw triaxial accelerometer data from Samsung wrist-worn fitness trackers. To anchor those data in the free-living environment, we compared the measurements with the physical activity scale of the Medlo behavioral observation scheme.
Results: We showed that consumer wearables are a valid tool for long-term data collection in this vulnerable patient population.
Conclusions: Regarding the movement intensity, the data collected by fitness trackers is overall in accordance with the data collected with the observational tool. Regarding the type of movement, we concluded that the automatic activity classification on the wearables is not yet ready for use with a mostly sedentary patient population.
Confocal laser endomicroscopy (CLE) enables real-time diagnosis of oral cancer and potentially malignant disorders by in vivo microscopic tissue examination. One impediment to the widespread clinical adoption of this technology is the need for operator expertise in image interpretation. Here we review the application of AI to automatic tissue classification of CLE images and discuss the opportunities for integrating this technology to advance the adoption of real-time digital pathology thus improving speed, precision and reproducibility.
This study aims to assess a new mobile application (app)’s efficacy in raising oral cancer awareness among older adults through educational videos and serious games. The app, named TEGO® (Tele-platform of Geriatric and Dental Specialties), with a video about oral cancer prevention, oral-self-examination, and serious gaming elements, like trivia and word search puzzles to reinforce the acquired knowledge was developed. Fifty-six patients, aged 60 to 80 years, were randomly selected from the Dental Clinic of the University of Chile and invited to use the app on their personal smartphones. Knowledge and attitudes were evaluated before two and four weeks after use. Oral self-examination practices were measured with a checkup guideline. The participation rate was 41.1%, mostly male (52.2%). Before using the app, 30.4% of the participants reported awareness of oral cancer, and none had performed oral self-examinations. Following two weeks after use, there was notable engagement, with 100% of participants utilizing it and responding that they had heard about oral cancer, and 56.5% having practiced an oral self-examination. This last outcome increased to 82.6% in the fourth week. The use of mHealth technologies has the potential as an effective educational tool for disseminating knowledge about oral cancer among older adults.
Cholera continues to pose a significant public health challenge in Nigeria, driven by poor sanitation, inadequate water quality, and climatic factors that create favorable conditions for outbreaks. Since the first epidemic in 1972, Nigeria has experienced recurrent outbreaks, with the most severe in 1991, resulting in over 7,000 deaths. Current surveillance systems and diagnostic methods are limited by infrastructural gaps, insufficient skilled personnel, and inadequate reporting, leading to delays in outbreak detection and response. These limitations exacerbate the public health burden, increasing mortality and the economic impact of cholera epidemics. This paper explores the potential of artificial intelligence (AI) and machine learning (ML) to address these challenges. AI technologies, including predictive modeling and ML algorithms such as random forests and convolutional neural networks (CNNs), can analyze diverse data sources-such as meteorological, environmental, and health records-to detect patterns and predict outbreaks. Case studies from other cholera-endemic regions, where AI achieved high predictive accuracy, demonstrate its transformative potential. By integrating AI into Nigeria’s public health infrastructure, early detection and response can be improved, resource allocation optimized, and disease transmission minimized. However, challenges such as data quality, standardization, and infrastructural deficits must be addressed. Multi-sectoral collaboration involving public health authorities, AI specialists, and policymakers is essential for the successful deployment of these technologies. This article concludes that AI-powered cholera surveillance systems have the potential to revolutionize public health outcomes, reducing cholera-related morbidity and mortality in resource-limited settings like Nigeria.
Atrial fibrillation (AF) is the most prevalent cardiac arrhythmia and can lead to severe complications such as stroke. Artificial intelligence (AI) has emerged as a vital tool in predicting and detecting AF, with machine learning (ML) models trained on electrocardiogram (ECG) data now capable of identifying high-risk patients or predicting the imminent onset of AF. Precision medicine aims to tailor medical interventions for specific sub-populations of patients who are most likely to benefit, utilizing large genomic datasets. Genetic studies have identified numerous loci associated with AF, yet translating this knowledge into clinical practice remains challenging. This paper explores the potential of AI in precision medicine for AF and examines its advantages, particularly when integrated with or compared to genomics. AI-driven ECG analysis provides a practical and cost-effective method for early detection and personalized treatment, complementing genomic approaches. AI-based diagnosis of AF allows for near-certain prediction, effectively relieving cardiologists of this task. In the context of preventive identification, AI enhances the accuracy of predictive models from 75% to 85% when ML is employed. In predicting the exact onset of AF-where human capability is virtually nonexistent-AI achieves a 74% accuracy rate, offering significant added value. The primary advantage of utilizing ECGs over genomic data lies in their ability to capture lifetime variations in a patient’s cardiac activity. AI-driven analysis of ECGs enables dynamic risk assessment and personalized adaptation of therapeutic strategies, optimizing patient outcomes. Genomics, on the other hand, enables the personalization of care for each patient. By integrating AI with ECG and genomic data, truly individualized care becomes achievable, surpassing the limitations of the “average patient” model.
Aim: Genetic instability represents the hallmark of carcinogenesis. For cancer, the retinoblastoma (RB) gene defect allowing genetic instability was successfully exploited to eliminate cancer. Similarly, this study aims to assess the genetic instability of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spike protein’s S1/S2 furin cleavage site in hopes of applying oligonucleotide-based therapeutics to suppress infectivity by exploiting hypermutability.
Methods: The Basic Local Alignment Search Tool was used to search for homology. Protein or nucleotide sequences were obtained from the National Center for Biotechnology Information database. BioEdit was used for multiple sequence alignment. Python-enhanced molecular graphics program was used for molecular modeling.
Results: To assess feasibility, comparative sequence alignment was performed on S1/S2 site plus juxtaposing residues of SARS-CoV-2 and avian infectious bronchitis virus (IBV) isolate AL/7052/97 that belongs to distinct genus. IBV amino acids correlating to 678-TNSPRRARSVASQS of SARS-CoV-2 spike protein were deciphered (nine identical, two conserved, two displaced, and one unconserved). The encoding nucleotides exhibited 14 identities, three transitions (C>U or U>C, two; G>A or A>G, one), and 15 transversions (U>A or A>U, eight; C>G or G>C, six; G>U or U>G, one) with mostly complementary base (14/15) for transversion. Analysis of SARS-CoV-2 variants corroborates that S1/S2 site continues to evolve. The overall data portrays an evolutionarily dynamic nature of S1/S2 site. The potential role of intragenomic ‘microhomology-mediated template switching’ by RNA-dependent RNA polymerase is described.
Conclusions: To apply virolytic pressure, peptide-guided oligonucleotides targeting S1/S2 site-encoding sequences may be deployed to trigger genomic RNA degradation. A potential consequence is that resistant variants (if emerge) may carry mutation(s) in S1/S2 site-encoding sequence to abrogate hybridization, which (by default) may encode defective substrate for furin. Thus, through ‘targeting oligonucleotides directed devolution’ of S1/S2 site, the infectivity of SARS-CoV-2 may be attenuated. An alternative strategy of oligonucleotide-based therapeutic editing by adenosine deaminases acting on RNA (ADAR) is mentioned.
Mental healthcare in a range of countries faces challenges, including rapidly increasing demand at a time of restricted access to services, insufficient mental healthcare professionals and limited funding. This can result in long delays and late diagnosis. The use of artificial intelligence (AI) technology to help to address these shortcomings is therefore being explored in a range of countries, including the UK. The recent increase in reported studies provides an opportunity to review the potential, benefits and drawbacks of this technology. Studies have included AI-based chatbots for patients with depression and anxiety symptoms; AI-facilitated approaches, including virtual reality applications in anxiety disorders; avatar therapy for patients with psychosis; AI humanoid robot-enhanced therapy, for both children and the isolated elderly in care settings; AI animal-like robots to help patients with dementia; and digital game interventions for young people with mental health conditions. Overall, the studies showed positive effects and none reported any adverse side effects. However, the quality of the data was low, mainly due to a lack of studies, a high risk of bias and the heterogeneity of the studies. Importantly also, longer-term effects were often not evident. This suggests that translating small-scale, short-term trials into effective large-scale, longer-term real-world applications may be a particular challenge. While the use of AI in mental healthcare appears to have potential, its use also raises important ethical and privacy concerns, potential risk of bias, and the risk of unintended consequences such as over-diagnosis or unnecessary treatment of normal emotional experiences. More robust, longer-term research with larger patient populations, and clear regulatory frameworks and ethical guidelines to ensure that patients’ rights, privacy and well-being are protected, are therefore needed.
Artificial intelligence (AI) increasingly influences biomedical scientific writing and clinical practice. The recent article by Fornalik et al. (Explor Digit Health Technol. 2024;2:235–48. doi: 10.37349/edht.2024.00024) explores AI’s capabilities, challenges, and ethical considerations in scientific communication, particularly highlighting tools like ChatGPT and Penelope.ai. This commentary aims to reflect on and expand the key themes presented by Fornalik et al. (Explor Digit Health Technol. 2024;2:235–48. doi: 10.37349/edht.2024.00024), emphasizing AI’s role in auditory healthcare, particularly in otolaryngology and auditory rehabilitation. The discussion is based on a critical review and synthesis of recent literature on AI applications in scientific writing and auditory healthcare. Key technologies such as generative AI platforms, machine learning algorithms, and mobile-based auditory training systems are highlighted. AI has shown promising results in enhancing manuscript preparation, literature synthesis, and peer review workflows. In clinical practice, adaptive AI models have improved cochlear implant programming, leading to up to 30% gains in speech perception accuracy. Mobile apps and telehealth platforms using AI have also improved listening effort, communication confidence, and access to care in remote settings. However, limitations include data privacy concerns, lack of population diversity in datasets, and the need for clinician oversight. AI presents transformative opportunities across biomedical science and healthcare. To ensure its responsible use, interdisciplinary collaboration among clinicians, researchers, ethicists, and technologists is essential. Such collaboration can help develop ethical frameworks that enhance innovation while safeguarding patient well-being and scientific integrity.
Background: The use of artificial intelligence (AI) has been shown to enhance human life quality by making it easier, safer, and more efficient. However, there is currently limited evidence about the applicability of AI in health insurance and easing the complexity of insurance operations. This study seeks to systematically review the literature related to the application, challenges, and opportunities of applying AI in the healthcare insurance industry.
Methods: A systematic review approach was utilized, guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The method included an exploratory and narrative design, a two-phase search strategy, eligibility criteria, and analysis.
Results: The search yielded 520 eligible articles. Twelve articles were eligible, evaluated, and analyzed in this study. Most articles discussed AI’s use in healthcare insurance to detect fraud, improve underwriting accuracy and transparency, and resolve medical information asymmetry. For claim processes, virtual agents, chatbots, customer engagement, telematics, and underwriting, algorithms were essential. However, technical skill is needed to create and deploy AI systems, and privacy was an issue due to massive data and algorithms that could abuse user data.
Discussion: The implementation of AI encounters various challenges, such as insufficient knowledge among users, a deficit in technical expertise and support, shortcomings in data strategy, and a growing reluctance towards AI. Privacy presents a challenge in AI, especially because of the widespread use of large data sets and algorithms that could misuse consumer information.
Aim: This study aimed to understand the mediating and moderating effects of self-esteem’s relationship with gaming disorder (GD).
Methods: Participants (N = 1,712) were recruited from online gaming forums. A battery of measures including GD, self-esteem, depression, anxiety, escapism, and playing time were completed.
Results: Escapism, depression, and playing time have a significant mediating effect on self-esteem’s relationship with GD. Escapism and depression explained most of the mediated effect, with playing time showing a much smaller effect. Anxiety was not a significant mediator. Unexpectedly, high self-esteem does not appear to buffer against the effects escapism and playing time have on GD. This contradicts clinical literature that promotes high self-esteem as a resilience factor.
Conclusions: Mediating effects of self-esteem’s relationship with GD were identified in this study. Moderators other than self-esteem might be more prudent to investigate in GD research.
This narrative review aims to appraise the evidence on artificial intelligence models for early diagnosis and risk stratification of oral cancer, focusing on data modalities, methodology differences, applications in the diagnostic flow and models’ performance. Models for early diagnosis and screening provide non-invasive diagnosis without the need for specialized instruments, which is ideal for early detection as a low-cost system. Supervised learning with well-annotated data provides reliable references for training the models, and therefore, reliable and promising results. Risk prediction models can be built based on medical record data, demographic data, clinical/histopathological descriptors, highly standardized images or a combination of these. Insights on which patients have a greater chance of malignancy development or disease recurrence can aid in providing personalized care, which can improve the patient’s prognosis. Artificial intelligence models demonstrate promising results in early diagnosis and risk stratification of oral cancer.
In recent years, patient engagement has emerged as a cornerstone in clinical decision-making, medical research, and health policy development, with its multifaceted value widely recognized by stakeholders across the healthcare continuum. However, digital health technologies, which are designed to enhance patient engagement, often fall short of their full potential due to developers’ limited understanding of patients’ needs and preferences. This perspective paper argues for adopting a patient-centered approach, emphasizing the critical importance of developers immersing themselves in patient communities to gain richer insights into patients’ lived experiences. Such an approach can lead to improved usability of digital health tools, enhanced user experience, and increased patient motivation, ultimately fostering more effective patient engagement in medical practice. Although challenges persist in the effective collection, analysis, and implementation of user feedback, prioritizing patient engagement remains crucial for optimizing health outcomes and enhancing the overall patient experience. By embracing this approach, developers can bridge the gap between technological innovation and patient needs, promoting more meaningful interactions and ultimately contributing to the advancement of healthcare systems and improved population health.
Background: The main theme of research literature on burnout has yet to be investigated. Aims: This bibliometric study evaluated the research literature on burnout and health, indexed in Web of Science (WoS), to reveal its expansion and the most prolific authors, institutions, countries, journals, and journal categories. The recurring themes of the literature were also identified.
Methods: In December 2023, the WoS Core Collection database was queried with: TS = [(“burnout*” OR “burn out*” OR “burn-out*”) AND (“health*” OR “illness*” OR “disease*” OR “well-being*” OR “wellbeing*”)]. The search yielded publications with these words presented in their title, abstract, or keywords. No filter was placed to restrict the search. Publication and citation counts were recorded directly from the database, whereas subsequent analyses were performed with VOSviewer.
Results: The search yielded 26,492 publications. The literature has been growing steadily in the 2000s and more quickly in the 2010s. Nearly one-third of the publications had contributions from the United States. The most prolific journals involved some open-access mega-journals and journals from psychology, medicine, and nursing. Depression and anxiety associated with burnout were recurring themes in the literature. The research community has been explaining burnout by the highly cited Job Demands-Resources (JD-R) model.
Discussion: This work demonstrated the usefulness of a bibliometric analysis to identify key stakeholders and major themes of burnout research.
The enormous global burden of cancer has created the need to develop cutting-edge strategies for enhancing public education on cancer. Over the years, conventional educational strategies, such as the use of posters and leaflets, have been preferentially employed as public education strategies on oral cancer; however, the use of digital education-based strategies has been largely underutilized. Notably, the use of digital education-based strategies, particularly serious games, has proven to be a superior cancer education strategy, when compared to conventional strategies, due to their rigorous design and features. This commentary discusses serious games as a game changer in cancer education, itemizing their diverse roles in cancer prevention, advocacy, and management. Also, this commentary also detailed those factors that might limit the use and availability of serious games in resource-limited settings.
Spatial memory, a fundamental cognitive function, enables individuals to encode, store, and retrieve information about their surroundings. Traditional assessment methods, such as paper-based tests and laboratory paradigms, often lack ecological validity and fail to capture the complexities of real-world navigation. Recent advancements in digital technologies, particularly virtual reality (VR) and mixed reality (MR), have introduced innovative tools for more immersive and accurate spatial memory assessments. VR provides controlled, replicable environments that simulate real-world navigation, while MR enhances engagement by blending virtual elements with physical spaces. This narrative review explores the cognitive mechanisms underlying spatial memory, highlighting the roles of egocentric and allocentric reference frames, as well as the neural substrates involved. The review also examines key factors influencing spatial memory performance, such as age, sex, neurological and neurodegenerative diseases. Digital tools such as the virtual Morris water maze and the VR Supermarket Test have been shown to possess enhanced ecological validity and diagnostic potential, particularly in the context of detecting early cognitive decline in Alzheimer’s disease. However, the field confronts several challenges, including the necessity for standardized protocols, the potential for adverse effects such as cybersickness, and the substantial cost associated with VR and MR systems. Future research directions in this field should include the integration of artificial intelligence for personalized assessments, and the combination of VR and MR tasks with neurophysiological techniques to advance understanding of spatial memory. Standardization, accessibility, and the creation of adaptive assessment for clinical populations will be crucial for optimizing the use of digital technologies in spatial memory research.
Aim: To explore patient satisfaction with telemedicine and its associated factors at the Rheumatology Outpatient Clinic, San Fernando Teaching Hospital (SFTH), and to determine patient preference for health-related consultations.
Methods: 305 patients were surveyed via consecutive sampling. Data was obtained via interviewer-administered questionnaires in a clinical setting, capturing demographics, challenges with face-to-face consultations, and patient perspectives on telemedicine. Items from the Telemedicine Satisfaction Questionnaire and Telehealth Usability Questionnaire were modified to capture impact. Data was analyzed using descriptive and inferential statistics (SPSS version 29).
Results: Most respondents were ≥ 40 years old (77.7%), Indo-Caribbean (66.2%), female (89.2%), unemployed (64.9%), and had secondary level education or higher (76.1%). Time off issues (13.0%), timing inconvenience (12.4%), and traveling costs (12.4%) were identified as challenges with face-to-face consultations. Fear of interaction (22.9%) and financial difficulty (22.7%), widely resulting from COVID-19, were additional challenges. Most patients reported satisfaction with telemedicine (71.5%), relating to easier access to health services (65.9%). Combined telemedicine and face-to-face consultations, as appropriate, were the most preferred option (73.4%). Several socio-demographic factors influenced patient satisfaction and preference for telemedicine services, with telemedicine convenience being the most significant factor.
Conclusions: The results conclude that patients at the Rheumatology Outpatient Clinic are satisfied with the current telemedicine service as a method of providing continuity of care (p > 0.001). Challenges encountered with face-to-face consultations and the COVID-19 pandemic can influence patients’ level of satisfaction with and preference for telemedicine. Telemedicine convenience was the most significant factor influencing patient satisfaction and preference (p > 0.001). Most patients’ preference for a combination approach of both telemedicine and face-to-face consultations reflects the current standard of care. The findings of this study suggest that telemedicine is reasonable to incorporate into outpatient care for patients with chronic rheumatological diseases.
Over the last four decades, lung cancer has been the leading cause of death in the United States. Non-small cell lung cancer (NSCLC) is the most common type of lung cancer, and historically, treatment consists of surgical resection, chemotherapy, and/or radiotherapy. Over the past decade, targeted immunotherapy has improved overall survival and treatment response. However, immunotherapy is expensive, and only select patients respond to immunotherapy. Recently, there has been much interest in using biomarkers to better identify and predict which patients will respond to therapy. There is much hope that the combined use of artificial intelligence (AI) and omics-based technology will provide enhanced capability to predict response to immunotherapy in patients with NSCLC. We performed a literature review and summarized the various approaches in which AI has been integrated with genomics, radiomics, pathomics, metabolomics, immunogenomics, and breathomics to better understand the tumor immune microenvironment and predict response to immunotherapy.
Aim: The spread of suicide and non-suicidal self-injury (NSSI) content on social media has raised ongoing concerns about user safety and mental health. In response, social media platforms like Twitter (now X) and Meta (i.e., Facebook and Instagram) introduced content moderation policies to mitigate harm and promote safer digital environments. This study explored immediate trends in user discourse surrounding suicide and NSSI following the enactment of Meta’s graphic self-harm imagery ban. Specifically, it examined shifts in tweet tone, content type, and underlying themes immediately before and after the policy’s implementation.
Methods: A corpus of 3,846 tweets was analyzed. Within this corpus, tweets spanning 32 weeks from October 18, 2018, to May 29, 2019, were selected. These dates were chosen to encompass approximately 16 weeks before and after the enactment of the policy on February 7, 2019. Tweets were categorized according to slant, tweet category, and theme.
Results: The findings revealed notable shifts in online discourse. There was a significant decrease in the proportion of tweets identified as anti-self-harm tweets and a corresponding increase in the proportion of tweets aimed at understanding self-harm, many of which were coded as personal opinions or informative content. These trends suggest that while content promoting self-harm did not increase, the tone of discourse shifted toward greater nuance and reflection. This may reflect users’ growing efforts to process, contextualize, and share perspectives on self-harm in a policy-regulated environment.
Conclusions: Meta’s graphic self-harm imagery ban appeared to influence how users communicated about suicide and NSSI on Twitter, prompting more content centered on understanding and discussion. However, the findings also highlight challenges in balancing harm reduction with space for personal narratives. These insights emphasize the role of policy in shaping public discourse and the need for clear moderation strategies that distinguish harmful promotion from lived experience and peer support.
Aim: This study aimed to evaluate virtual reality restorative environments (VRREs)’ impact on university students’ mental well-being, investigate factors affecting VRRE perception and psychological recovery, understand virtual environments’ healing mechanisms, and provide recommendations for university virtual healing spaces.
Methods: Semi-structured interviews were conducted with 32 participants to develop a VRRE perception evaluation system with five core and fourteen main categories. The system was then used to assess 13 virtual natural environments. Mental recovery effects were measured among 44 university students using the Schulte test (attention), Positive and Negative Affection Scale (mood), and physiological sensors (stress).
Results: VRREs demonstrated significant positive effects on participants’ psychological recovery. Different virtual environments showed varying impacts on attention, negative affect, and stress levels, while effects on positive affect were consistent across environments. Virtual extraterrestrial space environments yielded the strongest improvements in attention and stress reduction, whereas mixed forest settings were most effective in decreasing negative affects. Structural equation modeling revealed that participants’ VRRE perceptions significantly influenced psychological recovery through seven of fifteen pathways.
Conclusions: VRREs represent an effective intervention for supporting university students’ mental well-being. Different virtual environments offer distinct psychological benefits, with environment perception playing a crucial role in recovery outcomes. These findings provide valuable insights for designing targeted virtual healing spaces in university settings.
Aim: Structured light plethysmography (SLP), is a contactless optical system developed to monitor breathing patterns by analyzing chest-wall movement. It has not been thoroughly validated against other non-invasive motion analysis systems under different breathing conditions. This study therefore aimed to evaluate the criterion-validity of the SLP compared to the respiratory inductive plethysmography (RIP) at rest and after exercise.
Methods: Adults underwent two simultaneous 5-minute recordings from both devices, conducted at rest and following submaximal exercise on a cycle ergometer. Timing indices and thoracoabdominal (TA) movement parameters were examined. Measurement agreement between SLP and RIP was assessed using Bland-Altman plots at rest, after exercise, and for exercise-induced changes.
Results: Fifty adults (mean age 29.3 ± 6.8 years; 30 males) participated. Α total of 3,395 and 4,295 breath cycles were analyzed at rest and post-exercise, respectively. Over 92% of differences in timing parameters under both conditions were within the 95% limits of agreement (LOA) and their mean differences were found close to zero across a wide range of breath cycle magnitudes (rest: 2.62–8.06 s; post-exercise: 2.16–6.16 s). For ΤΑ movement parameters, the mean bias between devices at rest was 0.31 for ribcage amplitude (RCampi) and 0.23 for abdominal amplitude (ABampi), with LOA ranging from −0.06 to 0.66 and −0.06 to 0.52, respectively. A trend towards greater discrepancies for the individual measurements of RCampi and ABampi at higher magnitudes of TA movements was noted, especially post-exercise. A good average agreement between the devices was found for RCampi/ABampi both at rest [mean difference: 0.03, standard deviation (SD): 0.21] and after exercise (mean difference: 1.10, SD: 0.24).
Conclusions: The SLP is an accurate method to quantify and measure timing indices and the ratio of the ribcage motion to the abdominal motion under different breathing conditions.
Aim: Oxidative stress (OS) remains an intensively studied scientific problem. The quantitative measurement of OS is an unsolved task, largely due to the existence of numerous complex, non-linear interactions of its components, which can not be measured by traditional statistical methods. Modern mathematical processing based on artificial intelligence (AI) could be a promising method of OS assessment in medicine. The aim of the study was to investigate the potential possibilities of using multilayer neural networks to improve the diagnostic informativeness of the OS indicator-antioxidant (AO) activity (AOA) in patients with cardiovascular diseases (CVDs).
Methods: A cross-sectional study of a sample of 856 people, healthy volunteers and several groups of patients with CVDs (hypertension, including those complicated by coronary heart disease and/or cerebral ischemia, chronic cerebral ischemia), was carried out. The potentiometric method of determining the OS indicator, index of blood serum AOA, was used in comparison with a number of laboratory tests and clinical data. After the results of linear statistical evaluations were not satisfactory enough, а multilayer perceptron classifier was constructed for data analysis.
Results: By training a neural network, it was possible to assign a patient to one of the above-mentioned groups with 85% accuracy on the basis of 8 parameters selected from all the patients’ clinical and laboratory data, including the AOA value.
Conclusions: The use of multilayer neural networks can improve the diagnostic value of information obtained during the measurement of AOA index, in combination with simple laboratory tests in patients with CVDs. The application of AI algorithms is a promising tool to improve the laboratory measurement of OS and a potential solution to overcome the contradictions in the existing approaches to the evaluation of OS.
Aim: Colorectal cancer is a leading cause of cancer-related mortality, emphasising the need for accurate polyp segmentation during colonoscopy for early detection. Existing methods often struggle to generalize effectively across diverse clinical scenarios. This study introduces DeepPolyp, an artificial intelligence framework designed for comprehensive benchmarking and real-time clinical deployment of polyp segmentation models.
Methods: Transformer-based segmentation models, SegFormer and SSFormer, were trained from scratch using an extensive dataset comprising public collections (CVC-ClinicDB, ETIS-LaribPolypDB, Kvasir) and recently augmented datasets (PolypDataset-TCNoEndo, PolypGen). Training involved standardized data augmentation, learning rate schedules, and early stopping. Models were evaluated using Dice and Intersection over Union (IoU) metrics. Real-time inference performance was assessed on an NVIDIA Jetson Orin device with ONNX and TensorRT optimizations.
Results: SegFormer-B4 achieved the highest accuracy (Dice: 0.9843, IoU: 0.9694), but was not selected for clinical deployment due to computational constraints. SegFormer-B2 provided comparable accuracy (Dice: 0.9787, IoU: 0.9588) with significantly faster inference (94 ms per frame), offering an optimal balance suitable for real-time clinical use. SSFormer showed lower accuracy and slower inference, limiting its practical deployment.
Conclusions: DeepPolyp enables systematic evaluation of polyp segmentation models, assisting in selecting models based on both performance and computational efficiency. Despite superior accuracy from SegFormer-B4, SegFormer-B2 was selected for clinical deployment due to its advantageous balance between accuracy and real-time execution efficiency.
Aim: Moving beyond the traditional use/non-use dichotomy, this study examines how variations in older adults’ internet use relate to their multidimensional frailty status.
Methods: Data were drawn from the Belgian Ageing Studies (BAS), a large-scale cross-sectional survey conducted in Flanders (Belgium) and included 2,312 individuals aged 60 and older. Internet use was categorized into non-users, basic users, selective users and allround users. Multidimensional frailty was assessed using the Comprehensive Frailty Assessment Instrument (CFAI), covering physical, psychological, social and environmental domains. Multinomial logistic regression and Chi-squared automatic interaction detection (CHAID) were conducted.
Results: Regression analysis revealed that older adults with mild or high levels of physical frailty, as well as those with high levels of environmental frailty, were more likely to not use the internet. Furthermore, individuals with high physical frailty and high social frailty were more likely to be basic internet users. Social frailty was also linked to allround internet use, with those in the mild and high frailty categories being less likely to be allround users. However, CHAID analysis highlighted that sociodemographic factors-particularly low education and advanced age-are more strongly associated with low internet usage than frailty itself.
Conclusions: Multidimensional frailty is associated with internet use, with mild and high frailty groups being less internet savvy.
Virtual reality (VR) and digital health technologies have shown increasing potential in addressing psychological challenges such as homesickness and emotional distress, yet the role of emotional bonds, particularly place attachment, in shaping the design and effectiveness of these interventions remains underexplored. This study conceptualizes the integration of place attachment theory into digital health interventions, especially those utilizing VR, and proposes a theoretical and practical framework for designing emotionally resonant virtual environments. Two interrelated conceptual models are introduced: the Virtual Place Attachment Development Model (VPADM), which outlines psychological, social, environmental, and cultural dimensions that contribute to emotional bonding with virtual spaces, and the Cultural Adaptation System for Virtual Environments (CASVE), which addresses cross-cultural adaptation processes through assessment, implementation, and evaluation. These frameworks illustrate how virtual place attachment can be purposefully designed to enhance user engagement and emotional well-being, while also highlighting practical challenges such as accessibility, digital literacy, and the need for culturally responsive content. By integrating place attachment theory into digital mental health design, the paper offers a pathway to improve therapeutic outcomes in VR environments and provides a foundation for researchers and practitioners to develop emotionally supportive, culturally meaningful, and context-sensitive digital health interventions.
Digital mental health interventions (DMHIs) have demonstrated considerable potential to address mental health needs across diverse populations by offering scalable, adaptable, and cost-effective solutions. Nevertheless, their implementation in the Global South-particularly among Indigenous peoples and traditional communities-remains scattered and insufficiently systematised. These communities frequently face structural inequalities, limited access to formal mental health services, and distinct sociocultural frameworks that necessitate the development of culturally appropriate digital interventions. To date, existing reviews have focused predominantly on high-income countries, highlighting the urgent need to synthesise context-specific evidence from low- and middle-income settings. Therefore, this protocol outlines the design of a scoping review aimed at examining the available evidence on DMHIs targeting Indigenous peoples and traditional communities in the Global South. The scoping review will be conducted in accordance with the Joanna Briggs Institute (JBI) methodology for scoping reviews, and the results will be reported in line with the PRISMA-ScR guidelines. It will include articles resulting from primary research, systematic reviews, and opinion papers related to DMHIs in Indigenous and traditional populations in the Global South. No restrictions will be applied regarding the languages and year of publication. The search will be conducted in the following databases: MEDLINE, CINAHL, PsycINFO, Scopus, Embase, BVS Lilacs, African Index Medicus, and Index Medicus for the South-East Asia Region. Study selection and data extraction will be performed independently by three reviewers. The synthesis will include a numerical summary to provide an overview of the characteristics of the included studies, as well as a qualitative analysis aimed at identifying, analysing, and reporting recurring patterns and emerging categories within the data. The results will inform the development of future culturally competent digital mental health strategies tailored to the needs of Indigenous and traditional communities in the Global South.
Medicine is undergoing a deep technological transformation, with surgery on the cusp of this change, as technologies such as artificial intelligence (AI), augmented reality (AR), real-time imaging, and robotics converge to transform operative care. These innovations are now progressively being integrated into practice, driving precision surgery closer to reality. We aimed to assess how the convergence of AI, AR, real-time imaging, and robotics is advancing precision surgery and to outline the next wave of operative care. For doing this, we conducted a narrative perspective of publications that address AI-driven decision-making, AR-guided navigation, semi-autonomous robotics, and real-time imaging tailoring in surgical contexts. We observed that AI algorithms are expanding the potential of medicine by analyzing diverse data sets to optimize treatment strategies. AR-based navigation systems overlay digital anatomical information onto the surgical field, improving surgeon awareness and accuracy. Concurrently, AI-powered robotics are beginning to perform some surgical tasks semi-autonomously, potentially shortening procedure times and improving patient outcomes. Over time, the synergy between these disciplines may yield a new era of surgery: One where patient stratification guides operative decisions and where surgeons rely on data-driven systems for intraoperative feedback. This approach reinforces the principles of precision medicine and points toward a future in which surgery evolves hand in hand to improve clinical outcomes and patient safety. The synergy of data-driven surgery and personalized therapeutics brings a new era in precision medicine in which operative decisions may be dynamically tailored to the individual, promising greater safety and better clinical results.
Aim: In 2023, the average woman used 13 different personal care products (PCPs) daily, exposing them to 114 different chemical toxins, including carcinogens, synthetic preservatives, and fragrances. Parabens, commonly used as preservatives and fragrance ingredients, are found in household and PCPs, such as cosmetics and hair products. Exposure to parabens has been associated with an increased risk of breast cancer and endocrine disorders. This study aims to evaluate the effectiveness of the “Paraben-Free & Me” educational toolkit (OSF Registration DOI: 10.17605/OSF.IO/WXU34) by exploring whether it influences changes of 3.5 points (10%) in paraben-free behaviour in women aged 18–35 years old when compared to the control group.
Methods: This study consists of a randomized controlled trial to evaluate the effectiveness of the educational toolkit among 101 female students aged 18 to 35 years old following a four-week intervention period. Baseline and follow-up questionnaires were used to assess participants’ knowledge and access to information, risk perception, health beliefs, and paraben-free behaviour by providing composite scores for each construct.
Results: The change in the knowledge and access to information score was significantly greater in the intervention group compared to the control group (+3.96, 95% CI [1.95, 5.96]). There were slight differences between groups in relation to health beliefs (+0.63, 95% CI [–1.02, 2.27]), risk perception (+0.96, 95% CI [–0.61, 2.54]), and paraben-free behaviour (–0.16, 95% CI [–2.88, 2.57]); however, these were not statistically significant.
Conclusions: This study suggests that while the “Paraben-Free & Me” educational toolkit was unsuccessful in promoting greater paraben-free behaviour in the intervention group, it can increase women’s knowledge and access to information related to parabens in their PCPs. Future studies can focus on evaluating paraben-free behaviour and the effectiveness of the educational toolkit by exploring biological measures, such as urinary concentrations of parabens, and air pollutant concentrations.
Aim: Given the ubiquitous use of social media among young adults, understanding its impact on their psychological well-being is increasingly important. Research has identified negative associations between social media use and internalizing problems, such as depression and anxiety. Building on the previous research, the current study explored the mediating role of self-esteem in the associations between social media rumination (SMR) and symptoms of depression and anxiety in college students. Additionally, the study investigated the moderating role of gender in these associations.
Methods: The study sample consisted of 551 college students (mean age = 19 years; 36% men, 23.8% White) from a diverse midwestern university. The participants completed measures of depression (PHQ-9), anxiety (GAD-7), self-esteem (RSES), and SMR (Social Media Rumination Scale [SMRS]). An exploratory factor analysis was performed on the SMRS and supported a one-factor structure for the measure. Main analyses were conducted in R using PROCESS Model 4 and examined the associations between SMR and symptoms of depression and anxiety, with self-esteem as a mediator, and gender as a moderator. Additionally, time spent on social media and the number of posts per week were included as covariates in the analyses.
Results: Results indicated that SMR, above and beyond time spent on social media and type of engagement, was indirectly associated with depression and anxiety through self-esteem, and gender did not moderate these associations.
Conclusions: The study’s findings contribute to our understanding of the mechanisms linking social media use to internalizing problems, highlighting the crucial role of self-esteem in this process. Moreover, the study offers valuable insights for developing targeted interventions aimed at mitigating the negative effects of social media use on mental health by addressing SMR and bolstering self-esteem in young adults.
The prevalence of tinnitus in veterans is notably higher than in the general population and can significantly disrupt daily life. Given the impact of tinnitus, along with the lack of effective interventions, exploring new approaches is warranted. Wearable sound technologies offer a non-invasive and easily accessible approach. However, limited research has explored the effectiveness of sound therapy in UK veterans. A prior study supported the feasibility and acceptability of a non-invasive wearable device (i.e., TinniSoothe) in a sample of veterans. However, a waitlist-controlled trial is needed to investigate the effectiveness of the device. This waitlist-controlled trial aims to explore the effectiveness of a wearable device in reducing tinnitus symptoms in a sample of UK veterans. Veterans will be randomly allocated to one of two conditions: (1) the immediate intervention condition, which receives the device post-randomisation, or (2) the waitlist control group, which receives the device one-month post-randomisation. The trial will be conducted in veterans (n = 20) who have experienced tinnitus. Participants will be asked to use the device for one month. The immediate intervention group will be compared to the waitlist control group. The primary outcome is change in tinnitus severity (Tinnitus Functional Index, TFI) and mental health (General Health Questionnaire-12, GHQ-12) from baseline to one-month post-randomisation. Primary and secondary outcomes will be assessed at all timepoints (baseline, one-month post-randomisation, and two-month post-randomisation), while predictor variables will only be assessed at baseline to reduce participant burden. Recruitment will begin in October 2025. The study is expected to take 12 months, with results published in 2027. This study explores whether a wearable device is efficacious in reducing self-reported symptoms of tinnitus in comparison to a waitlist control group. This innovative approach, if successful, could offer a practical option for reducing tinnitus distress. The trial is registered on clinicaltrials.gov, identifier: NCT06905158.
The World Health Organization (WHO) estimates that unsafe food is responsible for 600 million cases and over 400,000 deaths annually. Traditional outbreak investigations are often time-consuming, inefficient, and limited by the quality and timeliness of available data. The integration of artificial intelligence (AI), such as machine learning, offers innovative approaches to improve the accuracy, speed, and efficiency of foodborne disease surveillance and outbreak detection. We conducted a mini review of the published literature and explored the potential applications of AI in foodborne disease prevention and control. Key areas explored included predictive analytics, food supply chain monitoring, public health surveillance, and laboratory-based investigations. AI-based predictive models support improved monitoring of environmental risk factors, better management of food supply chains, and more timely detection and prevention of contamination and outbreaks. We also described several challenges related to the integration of AI in food safety systems, including data quality, regulatory frameworks, and ethical considerations. By integrating advanced AI-driven methods, the future of food safety promises greater efficacy and equity in public health.
Aim: This study analyses the time series of daily increases in the number of diagnosed COVID-19 cases in Russia and countries from different continents. The aim of the study is to identify the specifics of the population response of different countries to the spread of the pandemic and anti-epidemic measures of public authorities to determine the most effective model to describe this process. This is a problematic, synoptic, and pilot study.
Methods: To evaluate this response strategy, models and methods from reliability theory are used to describe the probability of health protection, the probability density function of an increasing number of cases, the integrated risk of infection, the risk of morbidity, the acceptable risk, and the manageability of the epidemic situation. To approximate infection curves, various daily incidence rate functions are used and compared, and their coefficients are calculated for various pandemic waves.
Results: The results demonstrate that the Fréchet distribution function is the best model for the epidemic process. Indicators of variability in acceptable risk were identified during the first stage of pandemic development, showing the varying controllability of the situation by health systems. Through meta-analysis, country distributions were shown to appear as a single pattern, abstracted from local conditions. Estimated coefficients of reliability functions allow the construction of cartograms that reflect the peculiarities of state epidemic regulation and the stages of global pandemic deployment.
Conclusions: The findings confirm the effectiveness of the selected model in terms of reliability theory and identify directions for model improvement, taking into account the dynamic nature of the pandemic and its specific characteristics in different countries. The study is based on the methodological approach of function stratification (geometric fiber bundle). It allows for a deeper understanding of the identified patterns within a broader knowledge system.
Aim: Obsessive-compulsive disorder (OCD) is a mental health condition that significantly interferes with the school environment. The concealment of symptoms, lack of identification, and limited knowledge about the disorder often lead to delays in help-seeking, which are associated with greater chronicity, increased interference, and poorer treatment response. Programmes that educate teachers on early detection of OCD could help identify children at risk and promote help-seeking behavior. This study analyzed the feasibility, acceptability, and preliminary efficacy of the health app esTOCma from both quantitative and qualitative perspectives among teachers, as well as explored areas for improvement.
Methods: A total of 19 teachers (mean age = 47.74 years, SD = 11.2) completed the intervention along with pre- and post-intervention assessments through the app. In addition, they responded to open-ended questions to share their opinions about the app.
Results: Teachers took an average of 4.89 days (SD = 4.21) to complete the intervention. The app demonstrated excellent usability (M = 85.5, SD = 10.3) and was found useful by the majority of participants (89.5%), who reported satisfaction (84.2%) and stated they had learned considerably (73.7%) through its use. Upon completing the intervention, participants showed greater understanding of OCD and its treatments (MHLQ-R: z = –2.92, p = 0.004), lower levels of stigma (AQ-9: z = –3.67, p > 0.001), and a higher intention to seek professional help in case of experiencing obsessive-compulsive symptoms (GHSQ: z = –2.50, p = 0.012).
Conclusions: esTOCma appears to be a feasible app in an educational context, showing high acceptability among participating teachers. Moreover, the app increases knowledge and understanding of OCD, promotes the intention to seek professional help, and reduces stigma toward the disorder. Several improvements are suggested to further enhance the app’s potential impact in educational settings.
Aim: Designed to support the mental well-being of university students, the gamified Student Stress Resilience (SSResilience) app guides users in setting and working toward goals related to studying, socializing, and exercising. The app monitors progress through a combination of data from a user’s phone (via Internet of Things sensors) and information they enter themselves. This efficacy study documented students’ goal-setting efforts (RQ1) and examined the app’s effect on students’ anxiety, resilience, and psychological well-being (RQ2).
Methods: A quasi-experimental pretest-posttest control group design was used. Experimental group students (n1 = 25) used the app for two weeks. Control group students (n2 = 50) used different means to set the same goals. All students were pre-tested and post-tested on anxiety, well-being, and resilience using standardized questionnaires.
Results: Nineteen out of 25 experimental group students used the app to set one or more goals (19/25), and 18 of them found it helpful (18/25). The experimental group experienced a significant (t(22) = 2.72, P = 0.013) decrease in anxiety from Mpre = 8.96 (SD = 5.30) to Mpost = 5.76 (SD = 4.59), an increase in well-being from Mpre = 54.6 (SD = 25.88) to Mpost = 65.12 (SD = 23.90), but no change in resilience. Control group students’ (n = 43) measurements remained unchanged.
Conclusions: Preliminary findings indicate a potential value of the SSResilience app for significantly reducing students’ anxiety and increasing their well-being. Integrating Internet of Things technology (built-in phone sensors) into gamified apps for health holds significant promise by offering valuable data to users, app developers, and researchers. Future research will use wearables to measure stress and physical activity more accurately than self-reports.
We describe the rationale for and design of a non-inferiority trial to evaluate the relative effectiveness of electronic alcohol screening with in-person vs. electronic brief intervention (BI) approaches implemented in Alexandra Township, South Africa, and Zacatecas-Guadalupe, Mexico. The purpose of screening and brief intervention is to identify individuals whose responses to the Alcohol Use Disorders Identification Test (AUDIT) indicate risky drinking patterns and offer them information and advice to help them reduce their drinking. We seek to determine whether a BI comprising information and advice delivered electronically, along with the opportunity to schedule an appointment with a health care professional at a later time, is not significantly worse than a more labor-intensive traditional BI provided through a face-to-face interaction with a health professional immediately following screening. Selected patients visiting participating health clinics in Alexandra and Zacatecas-Guadalupe will be asked to complete the AUDIT screening using an online app accessed via a handheld device. Those whose scores indicate risky alcohol consumption will be invited to participate in the study. Participants at the clinics will be allocated in alternate weeks to either a customary in-person BI or an electronic BI. Based on power analyses taking attrition and nesting within clinics into account, the target sample sizes are 680 in Alexandra and 560 in Zacatecas-Guadalupe. Measures of 30-day alcohol consumption and AUDIT scores will be obtained at baseline, 3 months, and 6 months. The primary outcome will be the past 30-day quantity-frequency of alcohol consumption. Outcomes will be compared for the two study conditions using mixed effects multilevel regression analyses to account for nesting of observations within participants and participants within clinics. Potential socio-demographic covariates include gender, age, marital status, the highest completed level of education, family’s primary native language (a proxy for ethnicity/culture), presence of household members younger than 16, and subjective economic status (Trial ID: NCT07150156. Clinical trial platform: ClinicalTrials.gov Protocol Registration and Results System. Web address: https://clinicaltrials.gov).
Aim: Generative text-to-image technologies offer new opportunities for individuals to visually articulate internal experiences. While traditional artistic self-portraiture has been extensively associated with self-insight, its AI-assisted equivalent remains underexplored. This study investigates the experiential and assessment potential of AI-generated self-representations and assesses their applicability within contemporary digital mental health frameworks.
Methods: Five participants (aged 18–58) engaged in a 45-minute image generation session using Midjourney, producing approximately 500 images. This was followed by semi-structured interviews analyzed via interpretative phenomenological analysis (IPA). This study is idiographic and exploratory, drawing on the principles of IPA. Our aim is an in-depth explication of lived experience and shared experiential layers across cases. The findings should therefore be read as hypothesis-generating and as groundwork for future, larger-scale and mixed-methods evaluations and potential telepsychology integrations.
Results: Three group experiential themes: (1) images as functional tools (e.g., as sources of comfort or aspirational vision boards); (2) self-reflective space (facilitating spontaneous self-disclosure and novel insight); and (3) modalities of self-definition (symbolic representation and narrative arc). Participants described the process as highly engaging and reported enhanced self-efficacy.
Conclusions: AI-assisted image generation presents a flexible and user-centered modality for psychological reflection, with potential to augment art- and narrative-based therapeutic interventions. Ethical measures (e.g., anonymized data handling, withdrawal options) proved viable. Further research should explore larger, diverse samples and examine integration within telepsychology platforms to assess clinical utility.
Aim: To assess the burden of digital eye strain (DES) and associated factors among technology students at public universities in southern Ethiopia.
Methods: A cross-sectional study was conducted from March to April 2024 at three universities-Hawassa, Dilla, and Jinka. Data were collected using pretested self-administered questionnaires, including the Digital Eye Strain Questionnaire and other relevant variables.
Results: The survey tool was distributed to the total study population of 788 students, of whom 403 completed the survey, representing 93.5% of the calculated sample size of 431. Participants were predominantly male (74.7%) and enrolled at Hawassa University (57.8%). Daily digital device usage of ≥ 2 hours was reported by 259 (64.3%), and 72.7% had owned digital devices for > 2 years. Few participants reported smoking (0.7%), alcohol use (10.9%), or khat chewing (7.7%), and 13.9% had a history of accidents. Overall, 68.5% [95% confidence interval (CI): 64.0%–73.0%] experienced at least one symptom of DES in the past 12 months, with photophobia being the most common. DES was experienced more likely among students from Hawassa University [adjusted odds ratio (AOR) = 2.43; 95% CI: 1.11–5.30; p = 0.026], females (AOR = 2.32; 95% CI: 1.25–4.31; p = 0.008), current alcohol consumers (AOR = 3.12; 95% CI: 1.20–8.08; p = 0.019), and those with a history of accidents (AOR = 2.68; 95% CI: 1.17–6.13; p = 0.020).
Conclusions: Over two-thirds of final-year technology students in southern Ethiopian universities reported at least one symptom of DES, with higher risk observed among females, alcohol users, and those with prior accidents.
The integration of artificial intelligence (AI) into psychiatric care is rapidly revolutionizing diagnosis, risk stratification, therapy customization, and the delivery of mental health services. This narrative review synthesized recent research on ethical issues, methodological challenges, and practical applications of AI in psychiatry. A comprehensive literature search was conducted with no limitation to publication year using PubMed, Scopus, Web of Science, and Google Scholar to identify peer-reviewed articles and grey literature related to the integration of AI in psychiatry. AI enhances early identification, predicts relapses and treatment resistance, and facilitates precision pharmacopsychiatry by leveraging data from machine learning, natural language processing, digital phenotyping, and multimodal data integration. This review highlights the advancements in the integration of AI in psychiatric care, such as chatbot-mediated psychotherapy, reinforcement learning for clinical decision-making, and AI-driven triage systems in resource-constrained environments. However, there are still serious concerns about data privacy, algorithmic bias, informed consent, and the interpretability of AI systems. Other barriers to fair and safe implementation include discrepancies in training datasets, underrepresentation of marginalized groups, and a lack of clinician preparedness. There is a need for transparent, explainable, and ethically regulated AI systems that enhance, rather than replace, human decision-making. A hybrid human-AI approach to psychiatry is recommended to address these limitations, while interdisciplinary studies, strong validation frameworks, and inclusive policymaking are needed to guarantee that AI-enhanced mental health treatment continues to be effective, fair, and reliable.
Aim: Patients diagnosed with irregular astigmatism often require specific methods of vision correction. Among these, the use of a rigid gas permeable (RGP) lens is considered one of the most effective treatment approaches. This study aims to propose a new automated method for accurate RGP lens base-curve detection.
Methods: A multi-modal feature fusion approach was developed based on Pentacam images, incorporating image processing and machine learning techniques. Four types of features were extracted from the images and integrated through a serial feature fusion mechanism. The fused features were then evaluated using a multi-layered perceptron (MLP) network. Specifically, the features included: (1) middle-layer outputs of a convolutional autoencoder (CAE) applied to RGB map combinations; (2) ratios of colored areas in the front cornea map; (3) a feature vector from cornea front parameters; and (4) the radius of the reference sphere/ellipse in the front elevation map.
Results: Evaluations were performed on a manually labeled dataset. The proposed method achieved a mean squared error (MSE) of 0.005 and a coefficient of determination of 0.79, demonstrating improved accuracy compared to existing techniques.
Conclusions: The proposed multi-modal feature fusion technique provides a reliable and accurate solution for RGP lens base-curve detection. This approach reduces manual intervention in lens fitting and represents a significant step toward automated base-curve determination.