Free-form Text Research Articles

EDI&S SO04 Accuracy of Outsourced Radiology Reports in Emergency Surgical Care; Do they provide a high-quality cost-effective service?

Abstract Background Computed Tomography (CT) scans play a crucial role in emergency surgical care enabling proper management and prompt decision making. In our study we aimed to compare abdominal CT scan reports created by inhouse radiologists to those created by outsourced radiology reporting service for patients admitted with acute surgical pathologies. The study also looked at the impact of the reporting style (free text or structured format) on the quality of scan reports. Method The study was conducted in a busy general surgery unit located in the North West of England. Patients who were admitted acutely under the surgical specialty and had abdominal CT scans over a period of three months from August 2023 to October 2023 were included. Patients who were admitted following trauma or had non abdominal CT scans have been excluded. Statistical analysis was conducted to compare inhouse radiology reports to outsourced ones. Scan reports were thoroughly scrutinised to identify whether they included all relevant information needed and whether they were written in structured or free text format. Results A total of 192 patients were identified with a median age of 59 years. Outsourced radiology reports were found to be significantly underreporting comments on the biliary system (OR=0.369, P=0.046), spleen (OR=0.414, P=0.044), kidneys (OR=0.354, P=0.027), lymph nodes (OR=0.412, P=0.011), mesentery and peritoneum (OR=0.364, P=0.004), vasculature (OR=0.381, P=0.002), bones and soft tissues (OR=0.273, P=0.003), as well as lung bases (OR=0.300, P=0.009). Structured format was found to outperform free text style on comments on spleen (OR=8.064, P=0.044), adrenals (OR=8.468, P=0.039), mesentery and peritoneum (OR=4.618, P=0.000), as well as bone and soft tissues (OR=9.295, P=0.031). Conclusion CT scans reported by outsourced radiology service were found to be significantly underreporting certain key elements. Additionally, free style reporting format was found to miss some crucial aspects of abdominal scan reports. That adds further challenges to clinicians assessing and managing acute surgical patients. Further studies auditing different outsourced companies are recommended to ensure patients are provided with high-quality cost-effective emergency service.

Integrating AI in cause-of-death regular statistical production with human and expert system codings

Abstract Background Cause-of-death (CoD) statistics are key indicators in epidemiology and public health. These statistics are derived from death certificates completed by physicians and are usually coded by official statistics authorities according to the standards of the WHO ICD-10 classification to construct time- and cross-country-comparable statistics. For coding causes of death in free-text format in France according to ICD-10, predictions by deep neural networks (DNNs) are employed in addition to fully automatic batch coding by a rule-based expert system and to interactive coding by the coding team. Methods Seq-to-seq DNNs are trained from scratch on previously coded data to ICD-10 code multiple causes and underlying causes of death. Human coding is focused on certificates with a special public health interest or with low confidence in AI prediction quality to maximize the quality of the overall statistics disseminated when human resources are limited. DNNs also directly predict multiple causes and underlying causes of death for part of the certificates. Hence, the coding campaign aims to optimally allocate a coding mode to a given certificate. Results For deaths in 2021, 63% of the certificates are automatically batch coded by the expert system, 14% by the coding team, and 23% by DNNs. Compared to a traditional campaign that would have relied on automatic batch coding and manual coding, the present campaign reaches an accuracy of 95.7% for ICD-10 coding of the underlying cause (97.3% at the European shortlist level). Conclusions The 3-coding mode approach enables timeliness of dissemination of CoD statistics and may lead in the future to more responsive surveillance. Key messages • Deep neural networks trained on already coded data can be integrated into the regular statistical production of the CoD database in a fully controlled way. • The allocation between the different modes of coding can be optimized to achieve best quality and timeliness under the constraint of limited human resources.

GPT models classify free text from emergency department accesses in children with medical complexity

Abstract Background Emergency Department (ED) management software often collects clinical information as free text, but its unstructured nature limits routine use. While Machine Learning (ML) techniques have been employed to classify such information, they demand substantial data and computational resources. GPT-based models, using direct text prompts, offer a solution without requiring retraining, potentially simplifying clinical practice. Aims To assess OpenAI GPT’s performance in automating data extraction from ED medical records of complex pediatric patients. Methods Electronic admission records of children and young adults referring to the pediatric ED of Padova University Hospital (2007-2023) were analyzed. Principal variables of interest were discharge diagnosis, reason for ED access, required medical procedures, color code at triage, color code at discharge/hospitalization, outcome of the access. GPT-4 classification, via OpenAI API, was compared with manual classification by an expert pediatrician on sample of 791 records. Results Preliminary results of this study analyzed 107 records, showing a correct classification of discharge diagnosis in 95% of records. The reason for access to the ED was correctly identified in 97% of cases, while outcome of admission and color code, both at triage and discharge/hospitalization, were correctly classified in all given cases. Misclassification occurred mainly in cases where classification made by the model was incomplete (e.g., in the case of a trauma, abbreviated site of lesion was not identified by the model). Conclusions GPT-based models demonstrate feasibility, accuracy, and efficiency in transforming unstructured data into structured formats. Ultimately, GPT-based models could provide aid to the medical personnel in decision-making processes. Key messages • Emergency departments often collect clinical data in unstructured free text format, resulting in usability limitations. • GPT-powered models efficiently convert unstructured data to structured formats, offering potential support for medical decision-making.

Mutually-Guided Hierarchical Multi-Modal Feature Learning for Referring Image Segmentation

Referring image segmentation aims to locate and segment the target region based on a given textual expression query. The primary challenge is to understand semantics from visual and textual modalities and achieve alignment and matching. Prior works have attempted to address this challenge by leveraging separately pretrained unimodal models to extract global visual and textual features, and perform straightforwardly fusion to establish cross-modal semantic associations. However, these methods often concentrate solely on the global semantics, disregarding the hierarchical semantics of expression and image, and struggling with complex and open real scenarios, thus failing to capture critical cross-modal information. To address these limitations, this paper introduces an innovative mutually-guided hierarchical multi-modal feature learning scheme. By leveraging the guidance of global visual features, the model mines hierarchical text features from different stages of the text encoder. Simultaneously, the guidance of global textual features is leveraged to aggregate multi-scale visual features. This mutually guided hierarchical feature learning effectively addresses the semantically inaccurate cause by free-form text and naturally occurring scale variations. Furthermore, a Segment Detail Refinement (SDR) module is designed to enhance the model’s spatial detail awareness through attention mapping of low-level visual features and cross-modal features. To evaluate the effectiveness of the proposed approach, extensive experiments are conducted on three widely used referring image object segmentation datasets. The results demonstrate the superiority of the presented method in accurately locating and segmenting objects in images.

Strategies for Effective Product Roadmap Development and Execution in Data Analytics Platforms

Manufacturing is only one of several industries going through a digital transformation in this era of digital disruption. Manufacturing businesses are racing to adopt IoT-based solutions in order to innovate, increase productivity, lower costs, and gain greater market share because of the enormous transformational potential offered by Internet-of-Things (IoT) & Big Data. It is being used by industrial companies all over the world to increase flexibility while achieving cost savings, reduced inefficiencies, and better performance. It is no longer a trend for the future. But putting Industry 4.0 technology support into practice is a tough endeavour that gets increasingly harder in the absence of a common method. These are produced during the course of business operations and are kept in databases, email communication, transaction logs, free form texts on (business) social media, and other places. Businesses want to integrate data analytics methods into their decision-making processes by utilisation of these data. The field of Big Data analyses has seen tremendous progress in the IT sector in recent years. It appears that in order to deal with the extremely dynamic situations of today, new methods for product route mapping are required. This article provides an overview of the literature in science on product road mapping in order to shed light on the current state of the art & pinpoint research gaps. In order to demonstrate the influence of Industry 4.0 technological innovations on the manufacturing sector, this study provides an organised and content-focused evaluation of the literature. This paper offers recommendations for converting a legacy manufacturing facility into an Industry 4.0-compliant smart plant. Research gaps encompass topics including how to include objectives or results in product roadmaps, how to match a roadmap with the item's vision, and how to include activities related to product discovery in product roadmaps.

Development of message passing-based graph convolutional networks for classifying cancer pathology reports

BackgroundApplying graph convolutional networks (GCN) to the classification of free-form natural language texts leveraged by graph-of-words features (TextGCN) was studied and confirmed to be an effective means of describing complex natural language texts. However, the text classification models based on the TextGCN possess weaknesses in terms of memory consumption and model dissemination and distribution. In this paper, we present a fast message passing network (FastMPN), implementing a GCN with message passing architecture that provides versatility and flexibility by allowing trainable node embedding and edge weights, helping the GCN model find the better solution. We applied the FastMPN model to the task of clinical information extraction from cancer pathology reports, extracting the following six properties: main site, subsite, laterality, histology, behavior, and grade.ResultsWe evaluated the clinical task performance of the FastMPN models in terms of micro- and macro-averaged F1 scores. A comparison was performed with the multi-task convolutional neural network (MT-CNN) model. Results show that the FastMPN model is equivalent to or better than the MT-CNN.ConclusionsOur implementation revealed that our FastMPN model, which is based on the PyTorch platform, can train a large corpus (667,290 training samples) with 202,373 unique words in less than 3 minutes per epoch using one NVIDIA V100 hardware accelerator. Our experiments demonstrated that using this implementation, the clinical task performance scores of information extraction related to tumors from cancer pathology reports were highly competitive.

CURL Extensions – Weaving a Unique Property Reference Number into Address Histories in the Scottish Community Health Index

The Community Health Index (CHI) is the patient registration system for the National Health Service (NHS) in Scotland, covering the entire population. CHI has long been used as a population Indexing Spine to facilitate health-based data linkage for both operational and research purposes in Scotland. Changes in patient’s residential address since 2000 have been captured in a CHI Address History File. Addresses in CHI are recorded in unstructured free-text format, so property-based analyses and linkage is problematic. In August 2020, 90% of CHI addresses for currently registered patients, and people who had died that year, were seeded with a Unique Property Reference Number (UPRN). This product, called CHI-UPRN Residential Linkage (CURL), enabled NHS Scotland to harness important information on the patient population (5.8 million) and their place of residence, facilitating several data products related to COVID-19. The current project is extending CURL to UPRN-seed the 31 million address histories of everyone registered with a family doctor in Scotland since January 2000. The project, due for completion in Summer 2024, will look at alternative address-matching algorithms with the aim of achieving a 95% UPRN-match rate across the longitudinal CURL file. Quality assurance exercises help us to understand the precision and sensitivity of matching across time and geography. Analysis of how the longitudinal file is constructed across the person and property-level indices permits a deeper understanding of the capabilities and limitations of the CURL product when used to deliver innovative analyses or modelling that link people and places over time.

Lung Cancer Staging Using Chest CT and FDG PET/CT Free-Text Reports: Comparison Among Three ChatGPT Large-Language Models and Six Human Readers of Varying Experience.

Background: Although radiology reports are commonly used for lung cancer staging, this task can be challenging given radiologists' variable reporting styles as well as reports' potentially ambiguous and/or incomplete staging-related information. Objective: To compare performance of ChatGPT large-language models (LLMs) and human readers of varying experience in lung cancer staging using chest CT and FDG PET/CT free-text reports. Methods: This retrospective study included 700 patients (mean age, 73.8±29.5 years; 509 male, 191 female) from four institutions in Korea who underwent chest CT or FDG PET/CT for non-small cell lung cancer initial staging from January, 2020 to December, 2023. Examinations' reports used a free-text format, written exclusively in English or in mixed English and Korean. Two thoracic radiologists in consensus determined the overall stage group (IA, IB, IIA, IIB, IIIA, IIIB, IIIC, IVA, IVB) for each report using the AJCC 8th-edition staging system, establishing the reference standard. Three ChatGPT models (GPT-4o, GPT-4, GPT-3.5) determined an overall stage group for each report using a script-based application programming interface, zero-shot learning, and prompt incorporating a staging system summary. Six human readers (two fellowship-trained radiologists with lesser experience than the radiologists who determined the reference standard, two fellows, two residents) also independently determined overall stage groups. GPT-4o's overall accuracy for determining the correct stage among the nine groups was compared with that of the other LLMs and human readers using McNemar tests. Results: GPT-4o had an overall staging accuracy of 74.1%, significantly better than the accuracy of GPT-4 (70.1%, p=.02), GPT-3.5 (57.4%, p<.001), and resident 2 (65.7%, p<.001); significantly worse than the accuracy of fellowship-trained radiologist 1 (82.3%, p<.001) and fellowship-trained radiologist 2 (85.4%, p<.001); and not significantly different from the accuracy of fellow 1 (77.7%, p=.09), fellow 2 (75.6%, p=.53), and resident 1 (72.3%, p=.42). Conclusions: The best-performing model, GPT-4o, showed no significant difference in staging accuracy versus fellows, but significantly worse performance versus fellowship-trained radiologists. The findings do not support use of LLMs for lung cancer staging in place of expert healthcare professionals. Clinical Impact: The findings indicate the importance of domain expertise for performing complex specialized tasks such as cancer staging.

Changes in therapists’ knowledge of symptom management and stabilization following program co-participation with dissociative patients

Few professionals in the mental health field receive systematic training in treating trauma-related symptoms and disorders, including dissociative disorders (DD). Experts in the field of treating DDs recommend building stabilization skills early in treatment to improve emotion regulation and safety, yet research on DD therapists’ actual practices suggest that they are engaging in these practices less than recommended. DD patients may benefit from therapists learning more about emotion regulation and trauma symptom management towards stabilizing difficult experiences and risky, unhealthy, or unsafe behaviors. The current study considered whether an international group of therapists who participated alongside their DD patients in a 2-year Internet-based psychoeducational program demonstrated changes in knowledge related to symptom management and stabilization techniques. Therapists answered five free-form text questions related to DD symptom management and stabilization at baseline, mid-point, and end of the study. Results showed that significant changes in therapist knowledge were evident between baseline and mid-point, as well as between the baseline and the end of the study, with effect sizes ranging from small to large. Compared to when they began the study, therapists were better able to understand their DD patients’ reasons for self-injury, recognize warning signs of unsafe behaviors, identify coping skills, and identify skills to manage overwhelming feelings and intrusive traumatic content. Implications and opportunities for future research are discussed.

A vision-language foundation model for the generation of realistic chest X-ray images.

The paucity of high-quality medical imaging datasets could be mitigated by machine learning models that generate compositionally diverse images that faithfully represent medical concepts and pathologies. However, large vision-language models are trained on natural images, and the diversity distribution of the generated images substantially differs from that of medical images. Moreover, medical language involves specific and semantically rich vocabulary. Here we describe a domain-adaptation strategy for large vision-language models that overcomes distributional shifts. Specifically, by leveraging publicly available datasets of chest X-ray images and the corresponding radiology reports, we adapted a latent diffusion model pre-trained on pairs of natural images and text descriptors to generate diverse and visually plausible synthetic chest X-ray images (as confirmed by board-certified radiologists) whose appearance can be controlled with free-form medical text prompts. The domain-adaptation strategy for the text-conditioned synthesis of medical images can be used to augment training datasets and is a viable alternative to the sharing of real medical images for model training and fine-tuning.

Reshaping free-text radiology notes into structured reports with generative question answering transformers

BackgroundRadiology reports are typically written in a free-text format, making clinical information difficult to extract and use. Recently, the adoption of structured reporting (SR) has been recommended by various medical societies thanks to the advantages it offers, e.g. standardization, completeness, and information retrieval. We propose a pipeline to extract information from Italian free-text radiology reports that fits with the items of the reference SR registry proposed by a national society of interventional and medical radiology, focusing on CT staging of patients with lymphoma. MethodsOur work aims to leverage the potential of Natural Language Processing and Transformer-based models to deal with automatic SR registry filling. With the availability of 174 Italian radiology reports, we investigate a rule-free generative Question Answering approach based on the Italian-specific version of T5: IT5. To address information content discrepancies, we focus on the six most frequently filled items in the annotations made on the reports: three categorical (multichoice), one free-text (free-text), and two continuous numerical (factual). In the preprocessing phase, we encode also information that is not supposed to be entered. Two strategies (batch-truncation and ex-post combination) are implemented to comply with the IT5 context length limitations. Performance is evaluated in terms of strict accuracy, f1, and format accuracy, and compared with the widely used GPT-3.5 Large Language Model. Unlike multichoice and factual, free-text answers do not have 1-to-1 correspondence with their reference annotations. For this reason, we collect human-expert feedback on the similarity between medical annotations and generated free-text answers, using a 5-point Likert scale questionnaire (evaluating the criteria of correctness and completeness). ResultsThe combination of fine-tuning and batch splitting allows IT5 ex-post combination to achieve notable results in terms of information extraction of different types of structured data, performing on par with GPT-3.5. Human-based assessment scores of free-text answers show a high correlation with the AI performance metrics f1 (Spearman's correlation coefficients>0.5, p-values<0.001) for both IT5 ex-post combination and GPT-3.5. The latter is better at generating plausible human-like statements, even if it systematically provides answers even when they are not supposed to be given. ConclusionsIn our experimental setting, a fine-tuned Transformer-based model with a modest number of parameters (i.e., IT5, 220 M) performs well as a clinical information extraction system for automatic SR registry filling task. It can extract information from more than one place in the report, elaborating it in a manner that complies with the response specifications provided by the SR registry (for multichoice and factual items), or that closely approximates the work of a human-expert (free-text items); with the ability to discern when an answer is supposed to be given or not to a user query.

TOMGPT: Reliable Text-Only Training Approach for Cost-Effective Multi-modal Large Language Model

Multi-modal large language models (MLLMs), such as GPT-4, exhibit great comprehension capabilities on human instruction, as well as zero-shot ability on new downstream multi-modal tasks. To integrate the different modalities within a unified embedding space, previous MLLMs attempted to conduct visual instruction tuning with massive and high-quality image-text pair data, which requires substantial costs in data collection and training resources. In this article, we propose TOMGPT (Text-Only training Multi-modal GPT), a cost-effective MLLM tuned solely on easily accessible text data with much fewer resources. Along with pre-trained visual-linguistic coupled modality space (e.g., CLIP and ALIGN model), a text-only training strategy is devised to further project the aligned multi-modal latent space to that of LLM, endowing the LLM with visual comprehension capabilities in an efficient manner. Instead of enormous image-text training data required by previous MLLMs, we find that TOMGPT can be well-tuned with fewer yet diverse GPT-generated free-form text data, as we establish the semantic connection between LLM and pre-trained vision-language model. A quantitative evaluation is conducted on both MME and LVLM, which are recently released and extensively utilized MLLM benchmarks. The experiments reveal that TOMGPT achieved reliable performance compared to numerous models trained on a large amount of image-text pair data. Case studies are also presented, demonstrating TOMGPT’s broad understanding and dialogue capabilities across diverse image categories.

V. I. ABAEV'S HISTORICAL-ETYMOLOGICAL DICTIONARY: ISSUES IN THE DEVELOPMENT OF A DIGITAL BILINGUAL EDITION

In this article, we discuss the conclusions of the project on digitizing and translating the Historical-Etymological dictionary of Ossetic by V.I. Abaev, which has been carried out by the authors since 2020. The entries of the original include information of several kinds: Russian translations of an Ossetic word, related examples, usage and etymological notes. The etymological notes are usually free-form texts with a large number of cited forms from other languages (cognates, loanword sources, etc.). We argue for maximal preservation and thorough encoding of this information when creating a digital database. The optimal format for such purposes is XML markup, more specifically its variant intended for annotating text documents, TEI (Text Encoding Initiative, https://tei-c.org/). For digitizing the dictionary, we created a modification of this markup language, a module for Oxygen XML Editor, and a set of CSS files that ensure maximum similarity between our layout and the original. The same combination of XML and CSS allows generating the print version of the dictionary using the Prince typesetting system (https://www. princexml.com/) without any additional processing. For ease of search and access, we suggest supplementing the XML sources with a relational database automatically created from the lexical entries. In our project, this database serves as the source for the online interface of the dictionary (https://ossetic.iranic.space), powered by the OnLex platform [Makarov et al. 2022]. Hence, the project combines the features of several database types, which fully corresponds to the multi-faceted character of the original dictionary. We hope that our experience can be used for other digital lexicographic projects.

USING NATURAL LANGUAGE PROCESSING (NLP) FOR CATEGORIZING PAPER TITLES FROM GOOGLE FORMS

&amp;lt;p&amp;gt;Modern data collection, storage, and processing rely on diverse techniques to handle various types of information, ranging from structured tables to free-form text. This paper explores the captivating application of Natural Language Processing (NLP) for categorizing titles from Google Forms or any other textual data. The process of training an NLP model will be demonstrated through a specific example. Just as we learn from our past experiences, NLP models need to be fed with relevant data and labels. This ensures accurate and efficient processing even when new titles are introduced. We will conclude with a fascinating demonstration of how NLP algorithms analyze the structure and meaning of titles. By identifying keywords and understanding the context, they can automatically classify titles into relevant categories. This dramatically simplifies data organization and analysis, empowering us to extract valuable insights faster.&amp;lt;/p&amp;gt;

Using language to identify a bladder pain component in women with Dysmenorrhoea-Related Pelvic Pain: A cross-sectional study.

Dysmenorrhoea-Related Pelvic Pain (DRPP) is a common condition, which may or may not include bladder-related symptoms. Primary health care practitioners (PHCP) rely heavily on language for diagnosis of DRPP-related conditions. However, there are no established pain descriptors to assist PHCP to determine whether an individual's DRPP may include a bladder component. To identify differences in the use of pain descriptors in women with DRPP with and without a co-existing bladder pain component, through an exploratory study of the language of pelvic pain in women. A cross-sectional online survey of Australian and New Zealand women (n = 750, ages 18-49) who have self-identified pelvic pain. Free text and predetermined pain descriptors used by women with a self-perceived bladder pain component (DRPPB+, n = 468) were compared to those without bladder pain (DRPPB-, n = 282). Statistical analysis included Pearson χ2, logistic regression and analysis of variance tests using StataCorp Stata Statistical Software combined with qualitative data from AntConc concordance software. Within free-form text, bloating (P = 0.014) and pressure (P = 0.031) were used more commonly to describe dysmenorrhoea in women with DRPPB+, while the word excruciating (P < 0.001) was more commonly used by women with DRPPB-. From a pre-determined list of descriptors, pounding (P < 0.001), tingling (P < 0.001), stabbing (P = 0.010), burning (P = 0.002) and cramping (P = 0.021) were more commonly used by women with DRPPB+, than women with DRPPB-. Systematic patterns of word use should encourage practitioners to further enquire about bladder symptoms that may co-exist with dysmenorrhoea. Knowledge of these words may be useful in targeting diagnostic and therapeutic interventions.

EXPRESS: Gender and Serious Drug Recalls: a Textual Sentiment Analysis of Drug Reviews on WebMD

Consumers taking prescription drugs have limited ability to ascertain drug quality before taking the drug. After drug use, however, consumers frequently report their personal experiences with prescription drugs on one of the world’s largest medical websites: WebMD. Drug reviews on WebMD are a potentially rich source of free-form text that can be utilized to inform firms, consumers, researchers, and the Food and Drug Administration (FDA) about the quality and safety of prescription drugs. Additionally, because men and women communicate in starkly different ways, the gender of the reviewer may play a key role in drug reviews signaling drug quality problems. We examine if drug review textual sentiment is associated with the hazard of a serious drug recall and whether this relationship varies depending on the gender of the reviewer. We analyze textual sentiment on drug reviews from WebMD along with 13 years of drug recall data using several hazard models. We find that the more negative the drug review sentiment, the greater the hazard of a serious recall on that drug. This relationship is completely explained by drug reviews written by females; reviews written by males have no explanatory power. Our findings are confirmed by numerous robustness checks. In post-hoc analysis, we explore possible mechanisms by comparing female and male adverse events on the recalled drugs in our study. Our contributions to gender diversity and drug quality literature leads to implications for the FDA, WebMD, and firms that manufacture prescription drugs.

Empowering healthcare professionals in West Africa-A feasibility study and qualitative assessment of a dietary screening tool to identify adults at high risk of hypertension.

Dietary risks significantly contribute to hypertension in West Africa. Food frequency questionnaires (FFQs) can provide valuable dietary assessment but require rigorous validation and careful design to facilitate usability. This study assessed the feasibility and interest of a dietary screening tool for identifying adults at high risk of hypertension in Nigeria. Fifty-eight (58) consenting adult patients with hypertension and their caregivers and 35 healthcare professionals from a single-centre Nigerian hospital were recruited to complete a 27-item FFQ at two-time points and three 24-hour recalls for comparison in a mixed method study employing both quantitative questionnaires and qualitative techniques to elicit free form text. Data analyses were conducted using R software version 4.3.1 and NVivo version 14. The trial was registered with ClinicalTrials.gov: NCT05973760. The mean age of patients was 42.6 ± 11.9 years, with an average SBP of 140.3 ± 29.8 mmHg and a BMI of 29.5 ± 7.1 Kg/m2. The adherence rate was 87.9%, and the mean completion time was 7:37 minutes. 96.6% of patients found the FFQ easy to complete, comprehensive, and valuable. A minority reported difficulty (3.4%), discomfort (10.3%), and proposed additional foods (6.9%). Healthcare professionals considered the dietary screening tool very important (82.9%) and expressed a willingness to adopt the tool, with some suggestions for clarification. Patients and healthcare professionals found the screening tool favourable for dietary counselling in hypertension care. The tailored dietary screening tool (FFQ) demonstrated promising feasibility for integration into clinical care as assessed by patients and healthcare professionals. Successful implementation may benefit from proactive time management and addressing training needs. This user-centred approach provided key insights to refine FFQ and set the foundation for ongoing validity testing and evaluation in clinical practice.

Comparative evaluation of a language model and human specialists in the application of European guidelines for the management of inflammatory bowel diseases and malignancies.

Society guidelines on colorectal dysplasia screening, surveillance, and endoscopic management in inflammatory bowel disease (IBD) are complex, and physician adherence to them is suboptimal. We aimed to evaluate the use of ChatGPT, a large language model, in generating accurate guideline-based recommendations for colorectal dysplasia screening, surveillance, and endoscopic management in IBD in line with European Crohn's and Colitis Organization (ECCO) guidelines. 30 clinical scenarios in the form of free text were prepared and presented to three separate sessions of ChatGPT and to eight gastroenterologists (four IBD specialists and four non-IBD gastroenterologists). Two additional IBD specialists subsequently assessed all responses provided by ChatGPT and the eight gastroenterologists, judging their accuracy according to ECCO guidelines. ChatGPT had a mean correct response rate of 87.8%. Among the eight gastroenterologists, the mean correct response rates were 85.8% for IBD experts and 89.2% for non-IBD experts. No statistically significant differences in accuracy were observed between ChatGPT and all gastroenterologists (P=0.95), or between ChatGPT and the IBD experts and non-IBD expert gastroenterologists, respectively (P=0.82). This study highlights the potential of language models in enhancing guideline adherence regarding colorectal dysplasia in IBD. Further investigation of additional resources and prospective evaluation in real-world settings are warranted.

AI-Based System Automates Textual Classification of Daily Drilling Reports

_ This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper OTC 32978, “Development and Implementation of an AI-Based System To Automate Textual Classification on Daily Drilling Reports,” by Stephan Perrout, Aliel F. Riente, and Guilherme S.F. Vanni, SPE, Petrobras, et al. The paper has not been peer reviewed. Copyright 2023 Offshore Technology Conference. Reproduced by permission. _ Structured daily drilling reports (DDRs) are a rich source of information that allows better planning, more-accurate risk analysis, and improved key performance indicators and contracts. However, such information is originally stored in a free-text and unstructured format, which becomes difficult for efficient data mining. With the advance of artificial intelligence (AI) technologies, particularly AI language models, applying such techniques over unstructured data has become critical to digital transformation. The complete paper presents an approach for automatic DDR classification that incorporates new techniques of AI. Introduction This work addresses the complex task of automatic classification of DDRs according to a newly proposed ontology. The ontology follows a hierarchical model that classifies actions into three or four levels depending on the intervention, considering drilling, completion, and abandonment. Each event has an ontology built and reviewed by experts in oil and gas. Classifying DDR constitutes a demanding task, and effectively exploiting AI-based models represents a promising solution. This work bridges the gap by proposing a classifier based on transformers along with recurrent neural networks (RNNs) to classify reported events described in unstructured text related to drilling, completion, and abandonment interventions. A large number of DDRs was used for training and validation of the proposed classifier, yielding promising results for key processes in the company. Neural-Network Techniques Bidirectional Long Short-Term Memory. Early neural-network models are characterized by inputs of fixed length. This is a drawback when working with texts, however, because sentences vary in their number of words. To overcome such an issue and to process data sequentially, RNNs were proposed. The RNNs are characterized by a set of parameters inherent from the early models plus an internal memory (a hidden or internal state) responsible for storing the context of the sequence being processed. Long short-term memory (LSTM) is a variation of RNN proposed to mitigate two problems: Information can be easily lost when processing very long sequences, and the gradient can become quite low because of the high number of mathematical operations performed during the processing while remaining far from reaching the threshold. LSTM consists of a set of parameters called the input gate, forget gate, and output gate that control information flow through the network. This set of additional parameters helps to maintain only what is important for the internal state of the network besides controlling the output. BiLSTM is a variant of LSTM that comprises two LSTMs. One processes texts from left to right, and the second one processes texts from right to left. This feature allows “future” elements to be part of the model’s decision process for “past” elements. The final classification is the combination of the output of both LSTMs.

Exploratory evaluation of spinal cord stimulation with dynamic pulse patterns: a promising approach to improve stimulation sensation, coverage of pain areas, and expected pain relief.

The societal burden of chronic pain and the contribution-in-part to the opioid crisis, is a strong motivation to improve and expand non-addictive treatments, including spinal cord stimulation (SCS). For several decades standard SCS has consisted in delivery of tonic pulses with static parameter settings in frequency, pulse width, and amplitude. These static parameters have limited ability to personalize the quality of paresthesia, the dermatomal coverage, and thus may affect SCS efficacy. Further, static settings may contribute to the build-up of tolerance or loss of efficacy of the therapy over time in some patients. We conducted an acute exploratory study to evaluate the effects of SCS using time-dynamic pulses as compared to time-static (conventional tonic) stimulation pulses, with the hypotheses that dynamic pulse SCS may enable beneficial tailoring of the sensation and the patient's expectation for better pain relief with SCS. During a single clinic visit, consented subjects undergoing a standard SCS trial had their implanted leads temporarily connected to an investigational external stimulator capable of delivering time-static and six categories of time-dynamic pulse sequences, each characterized by continuously varying a stimulation parameter. Study subjects provided several assessments while blinded to the stimulation pattern, including: drawing of paresthesia maps, descriptions of sensation, and ratings for comfort and helpfulness to pain relief. Even without optimization of the field location, a majority of subjects rated sensations from dynamic stimulation as better or equal to that of static stimulation for comfortableness and for helpfulness to pain relief. The initial data showed a gender and/or pain dermatomal location related preference to a stimulation pattern. In particular, female subjects and subjects with pain at higher dermatomes tended to rank the sensation from dynamic stimulation better. Dynamic stimulation produced greater pain coverage without optimization; in 70% (9/13) of subjects, maximal pain coverage was achieved with a dynamic stimulation pattern. There was also greater variety in the words used by patients to describe stimulation sensation in the free text and free form verbal descriptions associated with dynamic stimulation. With the same electrode configuration and comparable parameter settings, acute SCS using dynamic pulses produced more positive ratings, expanded paresthesia coverage, and greater variation in sensation as compared to SCS using static pulses, suggesting that dynamic stimulation has the potential to improve capabilities of SCS for the treatment of chronic pain. Further study is warranted. This study was registered at ClinicalTrials.gov under ID NCT02988713, November 2016 (URL: https://clinicaltrials.gov/ct2/show/NCT02988713).