Speech recognition for medical documentation: an analysis of time, cost efficiency and acceptance in a clinical setting

Background/Aims Medical documentation is an important and unavoidable part of a health professional's working day. However, the time required for medical documentation is often viewed negatively, particularly by clinicians with heavy workloads. Digital speech recognition has become more prevalent and is being used to optimise working time. This study evaluated the time and cost savings associated with speech recognition technology, and its potential for improving healthcare processes. Methods Clinicians were directly observed while completing medical documentation. A total of 313 samples were collected, of which 163 used speech recognition and 150 used typing methods. The time taken to complete the medical form, the error rate and error correction time were recorded. A survey was also completed by 31 clinicians to gauge their level of acceptance of speech recognition software for medical documentation. Two-sample t-tests and Mann–Whitney U tests were performed to determine statistical trends and significance. Results On average, medical documentation using speech recognition software took just 5.11 minutes to complete the form, compared to 8.9 minutes typing, representing significant time savings. The error rate was also found to be lower for speech recognition software. However, 55% of clinicians surveyed stated that they would prefer to type their notes rather than use speech recognition software and perceived the error rate of this software to be higher than typing. Conclusions The results showed that there are both temporal and financial advantages of speech recognition technology over text input for medical documentation. However, this technology had low levels of acceptance among staff, which could have implications for the uptake of this method.

Modern technologies and mobile apps in developing speaking skill

Speaking is one of the abilities that students must develop when studying English. Speaking is a necessary technique for communication. Improving pupils' speaking ability has long been a priority in the classroom. Various novel technologies are being developed to educate speaking skills in classrooms in the rapidly expanding twenty-first century. Technology is the means through which we may gain access to this updated environment. Technology has been viewed as a means of assisting pupils in improving language abilities such as speaking ability. The Internet, podcasts, video conferencing, movies, and voice recognition software are regarded as the most effective instruments for training public speaking. The purpose of this paper is to explore some of the current tools that are accessible to English instructors today to help second or foreign language learners improve their speaking skills.

364. Individualized Prognostics in COVID-19 Facilitated by Computer Recognition of Blood Leukocyte Subsets

Background To determine whether CBC differentials of COVID+ inpatients can predict, at admission, both maximum oxygen requirements (MOR) and 30-day mortality. Methods Based on an approved IRB protocol, CBC differentials from the first 3 days of hospitalization of 12 SARS CoV-2 infected patients were retrospectively extracted from hospital records and analyzed with a privately owned Pattern Recognition Software (PRS, US Patent 10,429,389 B2) previously validated in sepsis, HIV, and hantavirus infections. PRS partitions the data into subsets immunologically dissimilar from one another, although internally similar. Results Regardless of the angle considered, the classic analysis −which measured the percentages of lymphocytes, monocytes, and neutrophils− did not distinguish outcomes (A). In contrast, non-overlapping patterns generated by the PRS differentiated 3 (left, vertical, and right) groups of patients (B). One subset was only composed of survivors (B). The remaining subsets included the highest oxygenation requirements (B). At least two immunologically interpretable, multi-cellular indicators distinguished the 3 data subsets with statistically significant differences (C, p≤ 0.05). Survivors (the left subset) showed lower N/L and/or higher M/L ratios than non-survivors (the vertical subset, C).Therefore, PRS partitioned the data into subsets that displayed both biological and significant differences. Because it offers visually explicit information, clinicians do not require a specialized training to interpret PRS-generated results. CBCs vs. outcomes - Software-analyzed CBCs vs. outcomes Conclusion (1) Analysis of blood leukocyte data predicts MOR and 30-d mortality. (2) Real time PRS analysis facilitates personalized medical decisions. (3) PRS measures two dimensions rarely assessed: multi-cellularity and dynamics. (4) Even with very small datasets, PRS may achieve statistical significance. (5) Larger COVID+ infected cohort is being analyzed for potential commercialization. Disclosures Claudia R. Libertin, MD, Gilead (Grant/Research Support)

De tolk als respeaker: een kwestie van training

As a result of political decisions, all Dutch-spoken television programmes that are broadcast by the Flemish Broadcasting Company (VRT) should be provided with subtitles for the deaf and hard of hearing by the year 2010. In order to meet these high expectations, the VRT is constantly improving and changing its subtitling techniques, as are other broadcasting companies worldwide. One of the main areas of change concerns the technique of live subtitling, i. e. real time subtitling of live television programmes. This type of subtitling has definitely benefited from the use of modern speech recognition software. Live subtitling, therefore, requires not only technical skills but also excellent ‘respeaking’ skills that are reminiscent of the skills of conference interpreters. The central question in the first part of this pa- per is to what extent ‘re-speaking’ is related to simultaneous (and other forms of) interpreting: is a good interpreter automatically a good re- speaker? In the second part, I adopt a didactic point of view in order to investigate the teaching aspects of real time subtitling skills: the conclusion is that it is best included in an education programme for conference interpreters

The Death of the Data Subject

This paper situates the data privacy debate in the context of what I call the death of the data subject. My central claim is that concept of a rights-bearing data subject is being pulled in two contradictory directions at once, and that simultaneous attention to these is necessary to understand and resist the extractive practices of the data industry. Specifically, it is necessary to treat the problems facing the data subject structurally, rather than by narrowly attempting to vindicate its rights. On the one hand, the data industry argues that subjects of biometric identification lack legal standing to pursue claims in court, and Facebook recently denied that that its facial recognition software recognizes faces. On the other hand, industry takes consent to terms of service and arbitration clauses to create enforceable legal subject positions, while using promises of personalization to create a phenomenological subject that is unaware of the extent to which it is being manipulated. Data subjects thus have no legal existence when it is a matter of corporate liability, but legal accountability when it is a matter of their own liability. Successful reform should address the power asymmetries between individuals and data companies that enable this structural disempowerment.

Estimation of accuracy loss by training a deep-learning-based cell organelle recognition software using full dataset and a reduced dataset containing only subviral particle distribution information

In collaboration with the Institute of Virology, Philipps University, Marburg, a deep-learning-based method that recognizes and classifies cell organelles based on the distribution of subviral particles in fluorescence microscopy images of virus-infected cells has been further developed. In this work a method to recognize cell organelles by means of partial image information is extended. The focus is on investigating loss of accuracy by only providing information about subviral particles and not all cell organelles to an adopted Mask-R convolutional neural network. Our results show that the subviral particle distribution holds information about the cell morphology, thus making it possible to use it for cell organelle-labelling.

Evaluation of the Diet Tracking Smartphone Application Keenoa™: A Qualitative Analysis

Keenoa™ is a novel Canadian diet application (app) currently used by Canadian dietitians to collect diet-related data from clients. The goal of this study was to evaluate Keenoa™ based on user feedback and compare it to a conventional pen and paper method. One hundred and two participants were recruited and randomly assigned to record their diets using this application for 3 nonconsecutive days. Following this, participants were invited to complete an online “exit” survey. Seventy-two subjects responded, with 50 completing an open-ended question asking for general feedback about the app. Data were reviewed and 3 main themes emerged: strengths, challenges, and future recommendations. Strengths associated with the app consisted of picture recognition software, the additional commentary feature, and the overall pleasant data collection process. Challenges that were identified included inconsistencies with the barcode scanning features, the limited food database, time to enter food details, and software issues. Future recommendations included using a larger food database, pairing dietary intake with physical activity monitoring, and having accessible nutritional data. Despite these limitations, participants preferred using mobile apps to record diet compared with traditional written food diaries.