Structuring medication signeturs as a language regression task: comparison of zero- and few-shot GPT with fine-tuned models.

Importance: Electronic health record textual sources such as medication signeturs (sigs) contain valuable information that is not always available in structured form. Commonly processed through manual annotation, this repetitive and time-consuming task could be fully automated using large language models (LLMs). While most sigs include simple instructions, some include complex patterns. Objectives: We aimed to compare the performance of GPT-3.5 and GPT-4 with smaller fine-tuned models (ClinicalBERT, BlueBERT) in extracting the average daily dose of 2 immunomodulating medications with frequent complex sigs: hydroxychloroquine, and prednisone. Methods: Using manually annotated sigs as the gold standard, we compared the performance of these models in 702 hydroxychloroquine and 22 104 prednisone prescriptions. Results: GPT-4 vastly outperformed all other models for this task at any level of in-context learning. With 100 in-context examples, the model correctly annotates 94% of hydroxychloroquine and 95% of prednisone sigs to within 1 significant digit. Error analysis conducted by 2 additional manual annotators on annotator-model disagreements suggests that the vast majority of disagreements are model errors. Many model errors relate to ambiguous sigs on which there was also frequent annotator disagreement. Discussion: Paired with minimal manual annotation, GPT-4 achieved excellent performance for language regression of complex medication sigs and vastly outperforms GPT-3.5, ClinicalBERT, and BlueBERT. However, the number of in-context examples needed to reach maximum performance was similar to GPT-3.5. Conclusion: LLMs show great potential to rapidly extract structured data from sigs in no-code fashion for clinical and research applications.

Transcriptional profiles of pulmonary artery endothelial cells in pulmonary hypertension.

Pulmonary arterial hypertension (PAH) is characterized by endothelial cell (EC) dysfunction. There are no data from living patients to inform whether differential gene expression of pulmonary artery ECs (PAECs) can discern disease subtypes, progression and pathogenesis. We aimed to further validate our previously described method to propagate ECs from right heart catheter (RHC) balloon tips and to perform additional PAEC phenotyping. We performed bulk RNA sequencing of PAECs from RHC balloons. Using unsupervised dimensionality reduction and clustering we compared transcriptional signatures from PAH to controls and other forms of pulmonary hypertension. Select PAEC samples underwent single cell and population growth characterization and anoikis quantification. Fifty-four specimens were analyzed from 49 subjects. The transcriptome appeared stable over limited passages. Six genes involved in sex steroid signaling, metabolism, and oncogenesis were significantly upregulated in PAH subjects as compared to controls. Genes regulating BMP and Wnt signaling, oxidative stress and cellular metabolism were differentially expressed in PAH subjects. Changes in gene expression tracked with clinical events in PAH subjects with serial samples over time. Functional assays demonstrated enhanced replication competency and anoikis resistance. Our findings recapitulate fundamental biological processes of PAH and provide new evidence of a cancer-like phenotype in ECs from the central vasculature of PAH patients. This "cell biopsy" method may provide insight into patient and lung EC heterogeneity to advance precision medicine approaches in PAH.

A feasibility study on AI-controlled closed-loop electrical stimulation implants.

Miniaturized electrical stimulation (ES) implants show great promise in practice, but their real-time control by means of biophysical mechanistic algorithms is not feasible due to computational complexity. Here, we study the feasibility of more computationally efficient machine learning methods to control ES implants. For this, we estimate the normalized twitch force of the stimulated extensor digitorum longus muscle on n = 11 Wistar rats with intra- and cross-subject calibration. After 2000 training stimulations, we reach a mean absolute error of 0.03 in an intra-subject setting and 0.2 in a cross-subject setting with a random forest regressor. To the best of our knowledge, this work is the first experiment showing the feasibility of AI to simulate complex ES mechanistic models. However, the results of cross-subject training motivate more research on error reduction methods for this setting.

Delirium detection using wearable sensors and machine learning in patients with intracerebral hemorrhage.

Objective: Delirium is associated with worse outcomes in patients with stroke and neurocritical illness, but delirium detection in these patients can be challenging with existing screening tools. To address this gap, we aimed to develop and evaluate machine learning models that detect episodes of post-stroke delirium based on data from wearable activity monitors in conjunction with stroke-related clinical features. Design: Prospective observational cohort study. Setting: Neurocritical Care and Stroke Units at an academic medical center. Patients: We recruited 39 patients with moderate-to-severe acute intracerebral hemorrhage (ICH) and hemiparesis over a 1-year period [mean (SD) age 71.3 (12.20), 54% male, median (IQR) initial NIH Stroke Scale 14.5 (6), median (IQR) ICH score 2 (1)]. Measurements and main results: Each patient received daily assessments for delirium by an attending neurologist, while activity data were recorded throughout each patient's hospitalization using wrist-worn actigraph devices (on both paretic and non-paretic arms). We compared the predictive accuracy of Random Forest, SVM and XGBoost machine learning methods in classifying daily delirium status using clinical information alone and combined with actigraph data. Among our study cohort, 85% of patients (n = 33) had at least one delirium episode, while 71% of monitoring days (n = 209) were rated as days with delirium. Clinical information alone had a low accuracy in detecting delirium on a day-to-day basis [accuracy mean (SD) 62% (18%), F1 score mean (SD) 50% (17%)]. Prediction performance improved significantly (p < 0.001) with the addition of actigraph data [accuracy mean (SD) 74% (10%), F1 score 65% (10%)]. Among actigraphy features, night-time actigraph data were especially relevant for classification accuracy. Conclusions: We found that actigraphy in conjunction with machine learning models improves clinical detection of delirium in patients with stroke, thus paving the way to make actigraph-assisted predictions clinically actionable.

Neural text generation in regulatory medical writing.

Background: A steep increase in new drug applications has increased the overhead of writing technical documents such as medication guides. Natural language processing can contribute to reducing this burden. Objective: To generate medication guides from texts that relate to prescription drug labeling information. Materials and Methods: We collected official drug label information from the DailyMed website. We focused on drug labels containing medication guide sections to train and test our model. To construct our training dataset, we aligned "source" text from the document with similar "target" text from the medication guide using three families of alignment techniques: global, manual, and heuristic alignment. The resulting source-target pairs were provided as input to a Pointer Generator Network, an abstractive text summarization model. Results: Global alignment produced the lowest ROUGE scores and relatively poor qualitative results, as running the model frequently resulted in mode collapse. Manual alignment also resulted in mode collapse, albeit higher ROUGE scores than global alignment. Within the family of heuristic alignment approaches, we compared different methods and found BM25-based alignments to produce significantly better summaries (at least 6.8 ROUGE points above the other techniques). This alignment surpassed both the global and manual alignments in terms of ROUGE and qualitative scoring. Conclusion: The results of this study indicate that a heuristic approach to generating inputs for an abstractive summarization model increased ROUGE scores, compared to a global or manual approach when automatically generating biomedical text. Such methods hold the potential to significantly reduce the manual labor burden in medical writing and related disciplines.

Correction: A case control study to investigate differences in motor control between individuals with and without non-specific low back pain during standing.

[This corrects the article DOI: 10.1371/journal.pone.0234858.].

Towards Objective Quantification of Hand Tremors and Bradykinesia Using Contactless Sensors: A Systematic Review.

Assessing the progression of movement disorders such as Parkinson's Disease (PD) is key in adjusting therapeutic interventions. However, current methods are still based on subjective factors such as visual observation, resulting in significant inter-rater variability on clinical scales such as UPDRS. Recent studies show the potential of sensor-based methods to address this limitation. The goal of this systematic review is to provide an up-to-date analysis of contactless sensor-based methods to estimate hand dexterity UPDRS scores in PD patients. Two hundred and twenty-four abstracts were screened and nine articles selected for analysis. Evidence obtained in a cumulative cohort of n = 187 patients and 1, 385 samples indicates that contactless sensors, particularly the Leap Motion Controller (LMC), can be used to assess UPDRS hand motor tasks 3.4, 3.5, 3.6, 3.15, and 3.17, although accuracy varies. Early evidence shows that sensor-based methods have clinical potential and might, after refinement, complement, or serve as a support to subjective assessment procedures. Given the nature of UPDRS assessment, future studies should observe whether LMC classification error falls within inter-rater variability for clinician-measured UPDRS scores to validate its clinical utility. Conversely, variables relevant to LMC classification such as power spectral densities or movement opening and closing speeds could set the basis for the design of more objective expert systems to assess hand dexterity in PD.

Estimating the COVID-19 Prevalence in Spain With Indirect Reporting via Open Surveys.

During the initial phases of the COVID-19 pandemic, accurate tracking has proven unfeasible. Initial estimation methods pointed toward case numbers that were much higher than officially reported. In the CoronaSurveys project, we have been addressing this issue using open online surveys with indirect reporting. We compare our estimates with the results of a serology study for Spain, obtaining high correlations (R squared 0.89). In our view, these results strongly support the idea of using open surveys with indirect reporting as a method to broadly sense the progress of a pandemic.

Predicting functional performance via classification of lower extremity strength in older adults with exergame-collected data.

OBJECTIVE: The goal of this article is to present and to evaluate a sensor-based functional performance monitoring system. The system consists of an array of Wii Balance Boards (WBB) and an exergame that estimates whether the player can maintain physical independence, comparing the results with the 30 s Chair-Stand Test (30CST). METHODS: Sixteen participants recruited at a nursing home performed the 30CST and then played the exergame described here as often as desired during a period of 2 weeks. For each session, features related to walking and standing on the WBBs while playing the exergame were collected. Different classifier algorithms were used to predict the result of the 30CST on a binary basis as able or unable to maintain physical independence. RESULTS: By using a Logistic Model Tree, we achieved a maximum accuracy of 91% when estimating whether player's 30CST scores were over or under a threshold of 12 points, our findings suggest that predicting age- and sex-adjusted cutoff scores is feasible. CONCLUSION: An array of WBBs seems to be a viable solution to estimate lower extremity strength and thereby functional performance in a non-invasive and continuous manner. This study provides proof of concept supporting the use of exergames to identify and monitor elderly subjects at risk of losing physical independence.

A Survey of Full-Body Motion Reconstruction in Immersive Virtual Reality Applications.

Due to recent advances in virtual reality (VR) technology, the development of immersive VR applications that track body motions and visualize a full-body avatar is attracting increasing research interest. This paper reviews related research to gather and to critically analyze recent improvements regarding the potential of full-body motion reconstruction in VR applications. We conducted a systematic literature search, matching VR and full-body tracking related keywords on IEEE Xplore, PubMed, ACM, and Scopus. Fifty-three publications were included and assigned in three groups: studies using markerless and marker-based motion tracking systems as well as systems using inertial measurement units. All analyzed research publications track the motions of the user wearing a head-mounted display and visualize a full-body avatar. The analysis confirmed that a full-body avatar can enhance the sense of embodiment and can improve the immersion within the VR. The results indicated that the Kinect device is still the most frequently used sensor (27 out of 53). Furthermore, there is a trend to track the movements of multiple users simultaneously. Many studies that enable multiplayer mode in VR use marker-based systems (7 out of 17) because they are much more robust and can accurately track full-body movements of multiple users in real-time.

Development of a Classifier to Determine Factors Causing Cybersickness in Virtual Reality Environments.

Objective: The goal of this contribution is to develop a classifier able to determine if cybersickness (CS) has occurred after immersion in a virtual reality (VR) scenario, based on a combination of biosignals and game parameters. Methods: We collected electrocardiographic, electrooculographic, respiratory, and skin conductivity data from a total of 66 participants. In addition, we also captured relevant game parameters such as avatar linear and angular speed as well as acceleration, head movements, and on-screen collisions. The data were collected while the participants were in a 10-minute VR experience, which was developed in Unity. The experience forced rotation and lateral movements upon the participants to provoke CS. A baseline was captured during a first simple scenario. The data were then split in per-level, per-60-second, and per-30-second windows. Furthermore, participants filled a pre- and postimmersion simulator sickness questionnaire. Simulator sickness scores were then used as a reference for binary (CS vs. no CS) and ternary (no CS-mild CS-severe CS) classification patterns. Several classification methods (support vector machines, K-nearest neighbors, and neural networks) were tested. Results: A maximum classification accuracy of 82% was achieved for binary classification and 56% for ternary classification. Conclusion: Given the sample size and the variety of movement patterns presented in the demonstration, we conclude that a combination of biosignals and game parameters suffice to determine the occurrence of CS. However, substantial further research is required to improve binary classification accuracy to adequate values for real-life scenarios and to determine better approaches to classify its severity.

Recent advances in rehabilitation for Parkinson's Disease with Exergames: A Systematic Review.

OBJECTIVE: The goal of this contribution is to gather and to critically analyze recent evidence regarding the potential of exergaming for Parkinson's disease (PD) rehabilitation and to provide an up-to-date analysis of the current state of studies on exergame-based therapy in PD patients. METHODS: We performed our search based on the conclusions of a previous systematic review published in 2014. Inclusion criteria were articles published in the indexed databases Pubmed, Scopus, Sciencedirect, IEEE and Cochrane published since January 1, 2014. Exclusion criteria were papers with a target group other than PD patients exclusively, or contributions not based on exergames. Sixty-four publications out of 525 matches were selected. RESULTS: The analysis of the 64 selected publications confirmed the putative improvement in motor skills suggested by the results of the previous review. The reliability and safety of both Microsoft Kinect and Wii Balance Board in the proposed scenarios was further confirmed by several recent studies. Clinical trials present better (n = 5) or similar (n = 3) results than control groups (traditional rehabilitation or regular exercise) in motor (TUG, BBS) and cognitive (attention, alertness, working memory, executive function), thus emphasizing the potential of exergames in PD. Pilot studies (n = 11) stated the safety and feasibility of both Microsoft Kinect and Wii Balance Board, potentially in home scenarios as well. Technical papers (n = 30) stated the reliability of balance and gait data captured by both devices. Related meta-analyses and systematic reviews (n = 15) further support these statements, generally citing the need for adaptation to patient's skills and new input devices and sensors as identified gaps. CONCLUSION: Recent evidence indicates exergame-based therapy has been widely proven to be feasible, safe, and at least as effective as traditional PD rehabilitation. Further insight into new sensors, best practices and different cognitive stadiums of PD (such as PD with Mild Cognitive Impairment), as well as task specificity, are required. Also, studies linking game parameters and results with traditional assessment methods, such as UPDRS scores, are required. Outcomes for randomized controlled trials (RCTs) should be standardized, and follow-up studies are required, particularly for motor outcomes.

Adapting smartphone-based photoplethysmograpy to suboptimal scenarios.

Photoplethysmography (PPG) is an optical technique used to measure the heart rate (HR) and other cardiovascular variables by analyzing volume changes in the microvascular bed of tissue. At the moment, smartphone users can already measure their HR using PPG applications that use the smartphone's built-in camera. However, available applications are unreliable when artifacts are present, such as those caused by movement, finger pressure, or ambient light changes. This contribution aims to analyze the limitations of a smartphone-based PPG algorithm capable of measuring N-N intervals when such artifacts are present by comparing it to a 2-lead electrocardiography (ECG). By using a Bandpass filter and a zero-crossing detection algorithm on a PPG signal captured at 800 × 600 pixels and 30 Hz, we have designed an approach capable of assessing N-N intervals when movement artifacts are present. An evaluation performed on n = 31 users shows our algorithm is capable of measuring N-N intervals with an average relative error of 9.23 ms, when compared to a 2-lead ECG. Our approach proves the reliability of smartphone-based photoplethysmography to measure N-N intervals, even under the presence of movement artifacts, and opens the door for its future use in remote diagnosis scenarios.

Cytochrome P450 CYP2B6 genotypes and haplotypes in a Colombian population: identification of novel variant CYP2B6 alleles.

OBJECTIVE: Information on CYP2B6 allele frequencies and detrimental genotypes in mixed human populations is scarce. The aim of this study was to analyze the frequencies and haplotypes of nonsynonymous CYP2B6 single nucleotide polymorphisms (SNPs) in a Colombian population. METHODS: One hundred and fifty-two healthy individuals were analyzed for five nonsynonymous CYP2B6 SNPs, namely rs8192709, rs3745274, rs2279343 rs28399499, and rs3211371. RESULTS: Besides eight known variant alleles, we identified two as yet unknown variant alleles combining, respectively, the SNPs rs3745274 and rs3211371 and rs8192709 and rs3745274. Comparison of Colombian mestizo individuals with other mestizo population indicates statistically significant differences (P<0.001) for the gain-of-function CYP2B6*4 allele and for combined detrimental CYP2B6 alleles. In addition, we observed a low linkage between the SNPs rs3745274 and rs2279343, which are often assumed as linked. CONCLUSION: In conclusion, large interethnic and intraethnic variability exists for CYP2B6 polymorphisms, thus reinforcing the need for tailored genotyping protocols for CYP2B6 testing as a biomarker of drug response.