Clinical evaluation of algorithms for context-sensitive physiological monitoring in children

Advances in pediatric perioperative care using artificial intelligence

PURPOSE OF THIS REVIEW

This article explores how artificial intelligence (AI) can be used to evaluate risks in pediatric perioperative care. It will also describe potential future applications of AI, such as models for airway device selection, controlling anesthetic depth and nociception during surgery, and contributing to the training of pediatric anesthesia providers.

RECENT FINDINGS

The use of AI in healthcare has increased in recent years, largely due to the accessibility of large datasets, such as those gathered from electronic health records. Although there has been less focus on pediatric anesthesia compared to adult anesthesia, research is on- going, especially for applications focused on risk factor identification for adverse perioperative events. Despite these advances, the lack of formal external validation or feasibility testing results in uncertainty surrounding the clinical applicability of these tools.

SUMMARY

The goal of using AI in pediatric anesthesia is to assist clinicians in providing safe and efficient care. Given that children are a vulnerable population, it is crucial to ensure that both clinicians and families have confidence in the clinical tools used to inform medical decision- making. While not yet a reality, the eventual incorporation of AI-based tools holds great potential to contribute to the safe and efficient care of our patients.

Use of artificial intelligence in paediatric anaesthesia: a systematic review

Objectives

Although the development of artificial intelligence (AI) technologies in medicine has been significant, their application to paediatric anaesthesia is not well characterised. As the paediatric operating room is a data-rich environment that requires critical clinical decision-making, this systematic review aims to characterise the current use of AI in paediatric anaesthesia and to identify barriers to the successful integration of such technologies.

Methods

This review was registered with PROSPERO (CRD42022304610), the international registry for systematic reviews. The search strategy was prepared by a librarian and run in five electronic databases (Embase, Medline, Central, Scopus, and Web of Science). Collected articles were screened by two reviewers. Included studies described the use of AI for paediatric anaesthesia (<18 yr old) within the perioperative setting.

Results

From 3313 records identified in the initial search, 40 were included in this review. Identified applications of AI were described for patient risk factor prediction (24 studies; 60%), anaesthetic depth estimation (2; 5%), anaesthetic medication/technique decision guidance (2; 5%), intubation assistance (1; 2.5%), airway device selection (3; 7.5%), physiological variable monitoring (6; 15%), and operating room scheduling (2; 5%). Multiple domains of AI were discussed including machine learning, computer vision, fuzzy logic, and natural language processing.

Conclusion

There is an emerging literature regarding applications of AI for paediatric anaesthesia, and their clinical integration holds potential for ultimately improving patient outcomes. However, multiple barriers to their clinical integration remain including a lack of high-quality input data, lack of external validation/evaluation, and unclear generalisability to diverse settings.

Systematic review protocol

CRD42022304610 (PROSPERO).

Artifact Processing Methods Influence on Intraoperative Hypotension Quantification and Outcome Effect Estimates

BACKGROUND

Physiologic data that is automatically collected during anesthesia is widely used for medical record keeping and clinical research. These data contain artifacts, which are not relevant in clinical care, but may influence research results. The aim of this study was to explore the effect of different methods of filtering and processing artifacts in anesthesiology data on study findings in order to demonstrate the importance of proper artifact filtering.

METHODS

The authors performed a systematic literature search to identify artifact filtering methods. Subsequently, these methods were applied to the data of anesthesia procedures with invasive blood pressure monitoring. Different hypotension measures were calculated (i.e., presence, duration, maximum deviation below threshold, and area under threshold) across different definitions (i.e., thresholds for mean arterial pressure of 50, 60, 65, 70 mmHg). These were then used to estimate the association with postoperative myocardial injury.

RESULTS

After screening 3,585 papers, the authors included 38 papers that reported artifact filtering methods. The authors applied eight of these methods to the data of 2,988 anesthesia procedures. The occurrence of hypotension (defined with a threshold of 50 mmHg) varied from 24% with a median filter of seven measurements to 55% without an artifact filtering method, and between 76 and 90% with a threshold of 65 mmHg. Standardized odds ratios for presence of hypotension ranged from 1.16 (95% CI, 1.07 to 1.26) to 1.24 (1.14 to 1.34) when hypotension was defined with a threshold of 50 mmHg. Similar variations in standardized odds ratios were found when applying methods to other hypotension measures and definitions.

CONCLUSIONS

The method of artifact filtering can have substantial effects on estimates of hypotension prevalence. The effect on the association between intraoperative hypotension and postoperative myocardial injury was relatively small. Nevertheless, the authors recommend that researchers carefully consider artifacts handling and report the methodology used.

Through the glass darkly: searching for safety signals in physiological monitoring

Perioperative monitoring of ventilation, gas exchange, heart rate, blood pressure, and other basic physiological measures give important information on the well-being of the child in the perioperative period. However, despite this level of surveillance, perioperative events that appear to be unheralded still occur. Improvements in alarms and alarm design combined with integrated analysis of monitored parameters that map to adverse outcomes may provide earlier warning of potential danger. Near real-time analysis of heart rate and blood pressure variability can provide information on autonomic function and cardiac reserve, while devices such as tissue oximetry may be beneficial to optimize regional and global blood flow.

Challenges of developing communicative interaction in individuals with congenital profound intellectual and multiple disabilities

BACKGROUND

Physiological responses have been used in individuals with acquired disability to enable communicative interaction without motor movement. This study explored four autonomic nervous system (ANS) signals-electrodermal activity, skin temperature, cardiac patterns and respiratory patterns-to enable interaction with individuals born with profound intellectual and multiple disabilities (PIMD).

METHOD

A series of case studies were conducted to teach a 15-year-old individual with PIMD to voluntarily control his ANS signals for the purposes of communicative interaction. Training was carried out according to an alternating treatment single-subject research design.

RESULTS

Training was unsuccessful; however, ANS signal patterns revealed issues unique to people with PIMD: profound intrasubject variability, discrepancy between physiological responses and caregiver perspectives, and the participant's lack of contingency awareness.

CONCLUSIONS

These three priority areas unique to people with congenital PIMD must be addressed before ANS signals can be used to enable communicative interaction with this population.

A vibro-tactile display for clinical monitoring: real-time evaluation

BACKGROUND

Vibro-tactile displays use human skin to convey information from physiological monitors to anesthesiologists, providing cues about changes in the status of the patient. In this investigation, we evaluated, in a real-time clinical environment, the usability and wearability of a novel vibro-tactile display belt recently developed by our group, and determined its accuracy in identifying events when used by anesthesiologists.

METHODS

A prospective observational study design was used. During routine anesthesia, a standard physiological monitor was connected to a software tool that used algorithms to automatically identify changing trends in mean noninvasive arterial blood pressure, expired minute ventilation, peak airway pressure, and end-tidal carbon dioxide partial pressure. The software was wirelessly interfaced to a vibro-tactile belt worn by the anesthesiologist. Each physiological variable was mapped to 1 of 4 tactor locations within the belt. The direction (increase/decrease) and 2 levels of change (small/large) were encoded in the stimulation patterns. A training session was completed by each anesthesiologist. The system was activated in real-time during anesthesia alongside routine physiological monitors. When the algorithms detected changes in the patient, the belt vibrated at the appropriate location with the pattern corresponding to the level and direction of change. Using a touch screen monitor the anesthesiologist was to enter the vibro-tactile message by first identifying the variable, then identifying the level and direction of change. Usability and wearability questionnaires were to be completed. The percentage of correct identification of the physiological trend, the direction of change, and the level of change were primary outcome variables. The mean usability score and wearability results were secondary outcome variables. We hypothesized that anesthesiologists would correctly identify the events communicated to them through the vibro-tactile belt 90% of the time, and that anesthesiologists would find the vibro-tactile belt usable and wearable.

RESULTS

Seventeen anesthesiologists evaluated the display during 57 cases. The belt was operational for a mean (SD) duration of 75 (41) minutes per case. Seven cases were excluded from analysis because of technical failures. Eighty-one percent (confidence interval [CI], 77% to 84%) of all stimuli were decoded. The physiological trend, the direction of change, and the level of change were correctly identified for 97.7% (CI 96%-99%), 94.9% (CI 92%-97%), and 93.5% of these stimuli (CI, 91%-96%), respectively. Fourteen anesthesiologists completed the usability and wearability questionnaires. The mean usability score was 4.8 of a maximum usability score of 7.

CONCLUSIONS

Anesthesiologists found a vibro-tactile belt to be wearable and usable and could accurately decode vibro-tactile messages in a real-time clinical environment.

An evaluation of the Integrated Pulmonary Index (IPI) for the detection of respiratory events in sedated patients undergoing colonoscopy

The Integrated Pulmonary Index (IPI™) is a new decision making tool calculated from measured end tidal carbon dioxide (etCO(2)), respiratory rate (RR), oxygen saturation (SpO(2)) and pulse rate (PR) using a fuzzy logic model. The aim of this study was to compare prospectively IPI to respiratory adverse events in patients undergoing moderate sedation for colonoscopy. Following ethics committee approval and personal informed consent 51 adult patients undergoing elective colonoscopy were enrolled. Patients received routine care by the endoscopy staff that were blinded to IPI, etCO(2), and RR; whilst a trained senior anesthesiologist observer, not involved in the procedure, collected this data. 'Requires attention' respiratory adverse events (at least 1 min of SpO(2) ≤ 92 % and/or RR ≤ 8 and or 20 % decrease in etCO(2)) and 'requires intervention' respiratory adverse events (at least 1 min of SpO(2) ≤ 85 % and/or RR = 0) were documented by the observer. There were no differences in etCO(2), RR, SpO(2) and PR between 5778 IPI readings ranging from 1 to 10. Low (1-3), medium (4-6) and high (7-10) IPI groups did not differ in RR, SpO(2) and PR, but etCO(2) was higher in the high IPI group (p = 0.0185). Among requires attention events (n = 113) the IPI was high (7-10) in 53.1 %, intermediate (4-6) in 32.7 %, or low (1-3) in 14.2 %. The presented data demonstrate limited agreement between respiratory physiological parameters and the IPI. Further IPI evaluation and validation is indicated mainly for events requiring immediate intervention and in different patient populations including obese patients.

A fuzzy logic-based system for anaesthesia monitoring

In recent years there has been a rapid growth in patient monitoring and medical data analysis using a number of computer-aided systems based on expert systems, fuzzy logic and many other intelligent techniques. Fuzzy logic-based expert systems have shown potential to improve clinician performance by imitating human thought processes in complex circumstances and accurately executing repetitive tasks to which humans are ill-suited. The main goal of this study was to develop a clinically useful diagnostic alarm system for detecting critical events during anaesthesia administration. The proposed diagnostic alarm system called Fuzzy logic monitoring system (FLMS) is presented. The performance of the system was validated through a series of off-line tests. When detecting hypovolaemia a substantial level of agreement was observed between FLMS and the human expert (the anaesthetist) during surgical procedures.