Applications and Performance of Machine Learning Algorithms in Emergency Medical Services: A Scoping Review.

OBJECTIVE: The aim of this study was to summarize the literature on the applications of machine learning (ML) and their performance in Emergency Medical Services (EMS). METHODS: Four relevant electronic databases were searched (from inception through January 2024) for all original studies that employed EMS-guided ML algorithms to enhance the clinical and operational performance of EMS. Two reviewers screened the retrieved studies and extracted relevant data from the included studies. The characteristics of included studies, employed ML algorithms, and their performance were quantitively described across primary domains and subdomains. RESULTS: This review included a total of 164 studies published from 2005 through 2024. Of those, 125 were clinical domain focused and 39 were operational. The characteristics of ML algorithms such as sample size, number and type of input features, and performance varied between and within domains and subdomains of applications. Clinical applications of ML algorithms involved triage or diagnosis classification (n = 62), treatment prediction (n = 12), or clinical outcome prediction (n = 50), mainly for out-of-hospital cardiac arrest/OHCA (n = 62), cardiovascular diseases/CVDs (n = 19), and trauma (n = 24). The performance of these ML algorithms varied, with a median area under the receiver operating characteristic curve (AUC) of 85.6%, accuracy of 88.1%, sensitivity of 86.05%, and specificity of 86.5%. Within the operational studies, the operational task of most ML algorithms was ambulance allocation (n = 21), followed by ambulance detection (n = 5), ambulance deployment (n = 5), route optimization (n = 5), and quality assurance (n = 3). The performance of all operational ML algorithms varied and had a median AUC of 96.1%, accuracy of 90.0%, sensitivity of 94.4%, and specificity of 87.7%. Generally, neural network and ensemble algorithms, to some degree, out-performed other ML algorithms. CONCLUSION: Triaging and managing different prehospital medical conditions and augmenting ambulance performance can be improved by ML algorithms. Future reports should focus on a specific clinical condition or operational task to improve the precision of the performance metrics of ML models.

Knowledge and willingness of schoolteachers in Jordan to perform CPR: a cross-sectional study.

OBJECTIVES: This study aimed to assess the cardiopulmonary resuscitation (CPR) knowledge and willingness of schoolteachers in Jordan. DESIGN: This was a cross-sectional study conducted using an online questionnaire. SETTING: For inclusion in this study, schoolteachers must be currently teaching at any level in schools across the country. Responses were collected from 1 April 2021 to 30 April 2021. PARTICIPANTS: All schoolteachers actively working in public or private schools were included in our study. PRIMARY AND SECONDARY OUTCOME MEASURES: Continuous variables were summarised as means and SD, whereas categorical variables were reported as frequencies and percentages (%). A χ2 test for independence, independent sample t-tests and analysis of variance were used appropriately. A p-value less than 0.05 was used to determine statistical significance. RESULTS: A total of 385 questionnaires were eligible for analyses. Only 14.5% of the participants received CPR training and overall correct knowledge answers were 18.8% of the total score. Those participants with previous CPR training had higher mean knowledge scores (2.34 vs 1.15, p<0.001). Trained participants were also more likely to provide hands-only CPR to various patient groups than untrained participants (p<0.05). Participants were more willing to provide standard CPR to family members than hands-only CPR (p<0.001), but more willing to provide hands-only CPR to friends (p<0.001), students (75.1% vs 58.2%, p<0.001), neighbour (p<0.001), stranger (p=0.001) and patient from the opposite gender (p<0.001). CONCLUSIONS: Schoolteachers in Jordan possess limited knowledge of CPR. However, the study participants showed a positive attitude towards performing CPR. The study revealed that they were more inclined to provide hands-only CPR than standard CPR. Policymakers and public health officials can take advantage of these findings to incorporate CPR training programmes for schoolteachers, either as a part of their undergraduate studies or as continuing education programmes with an emphasis on hands-only CPR.

Radiotherapy planning of spine and pelvis using single-energy metal artifact reduction corrected computed tomography sets.

Metal artifacts produce incorrect Hounsfield units and impact treatment planning accuracy. This work evaluates the use of single-energy metal artifact reduction (SEMAR) algorithm for treatment planning by comparison to manual artifact overriding. CT datasets of in-house 3D-printed spine and pelvic phantoms with and without metal insert(s) and two treated patients with metal implants were analysed. CT number accuracy improved with the use of SEMAR filter: root mean square deviation (RMSD) from reference (without metal) reduced by 35.4 in spine and 98.8 in hip. The plan dose volume histograms (DVHs) and dosimetric measurements showed comparable results. SEMAR reconstruction improved planning efficiency.

CyberKnife Xsight versus fiducial-based target-tracking: a novel 3D dosimetric comparison in a dynamic phantom.

BACKGROUND: The CyberKnife Xsight lung-tracking system (XLTS) provides an alternative to fiducial-based target-tracking systems (FTTS) for non-small-cell lung cancer (NSCLC) patients without invasive fiducial insertion procedures. This study provides a method for 3D independent dosimetric verification of the accuracy of the FTTS compared to the XLTS without relying on log-files generated by the CyberKnife system. METHODS: A respiratory motion trace was taken from a 4D-CT of a real lung cancer patient and applied to a modified QUASAR™ respiratory motion phantom. A novel approach to 3D dosimetry was developed using Gafchromic EBT3 film, allowing the 3D dose distribution delivered to the moving phantom to be reconstructed. Treatments were planned using the recommended margins for one and three fiducial markers and XLTS 2-view, 1-view and 0-view target-tracking modalities. The dose delivery accuracy was analysed by comparing the reconstructed dose distributions to the planned dose distributions using gamma index analysis. RESULTS: For the 3%/2 mm gamma criterion, gamma passing rates up to 99.37% were observed for the static deliveries. The 3-fiducial and 1-fiducial-based deliveries exhibited passing rates of 93.74% and 97.82%, respectively, in the absence of target rotation. When target rotation was considered, the passing rate for 1-fiducial tracking degraded to 91.24%. The passing rates observed for XLTS 2-view, 1-view and 0-view target-tracking were 92.78%, 96.22% and 76.08%, respectively. CONCLUSIONS: Except for the XLTS 0-view, the dosimetric accuracy of the XLTS was comparable to the FTTS under equivalent treatment conditions. This study gives us further confidence in the CyberKnife XLTS and FTTS systems.

Comparison of conventional versus customised Eurosil-4 Pink bolus for radiotherapy of the chest wall.

INTRODUCTION: In radiotherapy, the presence of air gaps near a tumour can lead to underdose to the tumour. In this study, the impact of air gaps on dose to the surface was evaluated. 3D-printing was used to construct a Eurosil-4 Pink bolus customised to the patient and its dosimetric properties were compared with that of Paraffin wax bolus. METHODS: Surface dose was measured for flat sheets of Eurosil-4 Pink bolus with different thicknesses. Different air gap thicknesses were inserted between the bolus and the surface, and dose was measured for each air gap using 10 cm × 10 cm fields. This was repeated with the effective field size calculated from the patient plan. Surface dose was measured for varying angles of incidence. A customised chest phantom was used to compare dose for two customised Eurosil-4 Pink boluses, and commonly used Paraffin wax bolus. RESULTS: The surface dose was found to be highest for 1.1 cm thick bolus. The decrease in surface dose for the Eurosil-4 Pink bolus was minimal for the 10 cm × 10 cm field, but higher for the effective field size and larger angles of incidence. For instance, the dose was reduced by 6.2% as a result of 1 cm air gap for the effective field size and 60 degree angle of incidence. The doses measured using Gafchromic film under the customised Eurosil-4 Pink boluses were similar to that of the Paraffin wax bolus, and higher than prescribed dose. CONCLUSIONS: The impact of air gaps can be significant for small field sizes and oblique beams. A customised Eurosil-4 Pink bolus has promising physical and dosimetric properties to ensure sufficient dose to the tumour, even for treatments where larger impact of air gaps is suspected.

Commissioning of a well type chamber for HDR and LDR brachytherapy applications: a review of methodology and outcomes.

For safe and accurate dose delivery in brachytherapy, associated equipment is subject to commissioning and ongoing quality assurance (QA). Many centres depend on the use of a well-type chamber ('well chamber') for performing brachytherapy dosimetry. Documentation of well chamber commissioning is scarce despite the important role the chamber plays in the whole brachytherapy QA process. An extensive and structured commissioning of the HDR 1000 plus well chamber (Standard Imaging Inc, Middleton WI) for HDR and LDR dosimetry was undertaken at Sir Charles Gairdner Hospital. The methodology and outcomes of this commissioning is documented and presented as a guideline to others involved in brachytherapy. The commissioning tests described include mechanical integrity, leakage current, directional dependence, response, length of uniform response, the influence of insert holders, ion collection efficiency, polarity effect, accuracy of measured air kerma strength (S(K)) or reference air kerma rate (K(R)) and baseline setting (for ongoing constancy checks). For the HDR 1000 plus well chamber, some of the insert holders modify the response curve. The measured sweet length was 2.5 cm which is within 0.5% of that specified by the manufacturer. Correction for polarity was negligible (0.9999) and ion recombination was small (0.9994). Directional dependence was small (less than 0.2%) and leakage current was negligible. The measured K(R) for (192)Ir agreed within 0.11% compared with a second well chamber of similar model and was within 0.5% of that determined via a free-in-air measurement method. Routine constancy checks over a year agreed with the baseline within 0.4%.