Trust in Machine Learning Driven Clinical Decision Support Tools Among Otolaryngologists.

BACKGROUND: Machine learning driven clinical decision support tools (ML-CDST) are on the verge of being integrated into clinical settings, including in Otolaryngology-Head & Neck Surgery. In this study, we investigated whether such CDST may influence otolaryngologists' diagnostic judgement. METHODS: Otolaryngologists were recruited virtually across the United States for this experiment on human-AI interaction. Participants were shown 12 different video-stroboscopic exams from patients with previously diagnosed laryngopharyngeal reflux or vocal fold paresis and asked to determine the presence of disease. They were then exposed to a random diagnosis purportedly resulting from an ML-CDST and given the opportunity to revise their diagnosis. The ML-CDST output was presented with no explanation, a general explanation, or a specific explanation of its logic. The ML-CDST impact on diagnostic judgement was assessed with McNemar's test. RESULTS: Forty-five participants were recruited. When participants reported less confidence (268 observations), they were significantly (p = 0.001) more likely to change their diagnostic judgement after exposure to ML-CDST output compared to when they reported more confidence (238 observations). Participants were more likely to change their diagnostic judgement when presented with a specific explanation of the CDST logic (p = 0.048). CONCLUSIONS: Our study suggests that otolaryngologists are susceptible to accepting ML-CDST diagnostic recommendations, especially when less confident. Otolaryngologists' trust in ML-CDST output is increased when accompanied with a specific explanation of its logic. LEVEL OF EVIDENCE: 2 Laryngoscope, 134:2799-2804, 2024.

A deep learning pipeline for automated classification of vocal fold polyps in flexible laryngoscopy.

PURPOSE: To develop and validate a deep learning model for distinguishing healthy vocal folds (HVF) and vocal fold polyps (VFP) on laryngoscopy videos, while demonstrating the ability of a previously developed informative frame classifier in facilitating deep learning development. METHODS: Following retrospective extraction of image frames from 52 HVF and 77 unilateral VFP videos, two researchers manually labeled each frame as informative or uninformative. A previously developed informative frame classifier was used to extract informative frames from the same video set. Both sets of videos were independently divided into training (60%), validation (20%), and test (20%) by patient. Machine-labeled frames were independently verified by two researchers to assess the precision of the informative frame classifier. Two models, pre-trained on ResNet18, were trained to classify frames as containing HVF or VFP. The accuracy of the polyp classifier trained on machine-labeled frames was compared to that of the classifier trained on human-labeled frames. The performance was measured by accuracy and area under the receiver operating characteristic curve (AUROC). RESULTS: When evaluated on a hold-out test set, the polyp classifier trained on machine-labeled frames achieved an accuracy of 85% and AUROC of 0.84, whereas the classifier trained on human-labeled frames achieved an accuracy of 69% and AUROC of 0.66. CONCLUSION: An accurate deep learning classifier for vocal fold polyp identification was developed and validated with the assistance of a peer-reviewed informative frame classifier for dataset assembly. The classifier trained on machine-labeled frames demonstrates improved performance compared to the classifier trained on human-labeled frames.

The use of thyroid isthmusectomy for management of well differentiated thyroid carcinoma - A systematic review and meta-analysis.

OBJECTIVE: With the growing global incidence of thyroid carcinomas, there is an increasing need for distinct guidelines for isthmus-confined carcinomas. Here, we performed the first systematic review on the topic to date, aiming to provide understanding to isthmusectomy as surgical management for well-differentiated thyroid carcinoma of the isthmus. METHODS: We conducted a systematic review following the PRISMA guidelines, analyzing English-language studies from the past decade that report on thyroid isthmusectomy. Exclusion criteria included isthmusectomy performed alongside full thyroidectomy or partial thyroid lobectomy, lack of data on tumor characteristics or survival outcomes, and non-English publications where a translation was unavailable. Our review identified a total of 227 patients from seven studies. RESULTS: The average 5-year overall survival and disease-free survival rates for patients with isthmus-confined PTC who underwent isthmusectomy were 100 % and 93.1 %, respectively. Similar to that of total thyroidectomy. 3.1 % of patients required completion thyroidectomy. Furthermore, isthmusectomy resulted in fewer surgical complications than total thyroidectomy. CONCLUSIONS: The scarcity of studies providing detailed tumor characteristics and patient outcomes limits our ability to fully evaluate the safety and efficacy of isthmusectomy for isthmus-confined PTC. Additionally, the variable sample sizes and restricted geographic distribution of the included studies calls into questions the generalizability of their findings. Despite these limitations, the data suggest that isthmusectomy may be a viable surgical option for select patients with small, isthmus-confined PTC. In the absence of a randomized controlled trial on the noninferiority of isthmusectomy, significantly more publications are needed before strong conclusions can be drawn.

Radiomic Analysis to Predict Histopathologically Confirmed Pseudoprogression in Glioblastoma Patients.

Purpose: Pseudoprogression mimicking recurrent glioblastoma remains a diagnostic challenge that may adversely confound or delay appropriate treatment or clinical trial enrollment. We sought to build a radiomic classifier to predict pseudoprogression in patients with primary isocitrate dehydrogenase wild type glioblastoma. Methods and Materials: We retrospectively examined a training cohort of 74 patients with isocitrate dehydrogenase wild type glioblastomas with brain magnetic resonance imaging including dynamic contrast enhanced T1 perfusion before resection of an enhancing lesion indeterminate for recurrent tumor or pseudoprogression. A recursive feature elimination random forest classifier was built using nested cross-validation without and with O6-methylguanine-DNA methyltransferase status to predict pseudoprogression. Results: A classifier constructed with cross-validation on the training cohort achieved an area under the receiver operating curve of 81% for predicting pseudoprogression. This was further improved to 89% with the addition of O6-methylguanine-DNA methyltransferase status into the classifier. Conclusions: Our results suggest that radiomic analysis of contrast T1-weighted images and magnetic resonance imaging perfusion images can assist the prompt diagnosis of pseudoprogression. Validation on external and independent data sets is necessary to verify these advanced analyses, which can be performed on routinely acquired clinical images and may help inform clinical treatment decisions.

Segmentation of Vestibular Schwannomas on Postoperative Gadolinium-Enhanced T1-Weighted and Noncontrast T2-Weighted Magnetic Resonance Imaging Using Deep Learning.

OBJECTIVE: Surveillance of postoperative vestibular schwannomas currently relies on manual segmentation and measurement of the tumor by content experts, which is both labor intensive and time consuming. We aimed to develop and validate deep learning models for automatic segmentation of postoperative vestibular schwannomas on gadolinium-enhanced T1-weighted magnetic resonance imaging (GdT1WI) and noncontrast high-resolution T2-weighted magnetic resonance imaging (HRT2WI). STUDY DESIGN: A supervised machine learning approach using a U-Net model was applied to segment magnetic resonance imaging images into pixels representing vestibular schwannoma and background pixels. SETTING: Tertiary care hospital. PATIENTS: Our retrospective data set consisted of 122 GdT1WI and 122 HRT2WI studies in 82 postoperative adult patients with a vestibular schwannoma treated with subtotal surgical resection between September 1, 2007, and April 17, 2018. Forty-nine percent of our cohort was female, the mean age at the time of surgery was 49.8 years, and the median time from surgery to follow-up scan was 2.26 years. INTERVENTIONS: N/A. MAIN OUTCOME MEASURES: Tumor areas were manually segmented in axial images and used as ground truth for training and evaluation of the model. We measured the Dice score of the predicted segmentation results in comparison to manual segmentations from experts to assess the model's accuracy. RESULTS: The GdT1WI model achieved a Dice score of 0.89, and the HRT2WI model achieved a Dice score of 0.85. CONCLUSION: We demonstrated that postoperative vestibular schwannomas can be accurately segmented on GdT1WI and HRT2WI without human intervention using deep learning. This artificial intelligence technology has the potential to improve the postoperative surveillance and management of patients with vestibular schwannomas.

A scoping review of the unassisted physical exam conducted over synchronous audio-video telemedicine.

BACKGROUND: This scoping review aims to provide a broad overview of the research on the unassisted virtual physical exam performed over synchronous audio-video telemedicine to identify gaps in knowledge and guide future research. METHODS: Searches for studies on the unassisted virtual physical exam were conducted in 3 databases. We included primary research studies in English on the virtual physical exam conducted via patient-to-provider synchronous, audio-video telemedicine in the absence of assistive technology or personnel. Screening and data extraction were performed by 2 independent reviewers. RESULTS: Seventy-four studies met inclusion criteria. The most common components of the physical exam performed over telemedicine were neurologic (38/74, 51%), musculoskeletal (10/74, 14%), multi-system (6/74, 8%), neuropsychologic (5/74, 7%), and skin (5/74, 7%). The majority of the literature focuses on the telemedicine physical exam in the adult population, with only 5% of studies conducted specifically in a pediatric population. During the telemedicine exam, the patients were most commonly located in outpatient offices (28/74, 38%) and homes and other non-clinical settings (25/74, 34%). Both patients and providers in the included studies most frequently used computers for the telemedicine encounter. CONCLUSIONS: Research evaluating the unassisted virtual physical exam is at an early stage of maturity and is skewed toward the neurologic, musculoskeletal, neuropsychologic, and skin exam components. Future research should focus on expanding the range of telemedicine exam maneuvers studied and evaluating the exam in the most relevant settings, which for telemedicine is trending toward exams conducted through mobile devices and in patients' homes.

Automatic classification of informative laryngoscopic images using deep learning.

Objective: This study aims to develop and validate a convolutional neural network (CNN)-based algorithm for automatic selection of informative frames in flexible laryngoscopic videos. The classifier has the potential to aid in the development of computer-aided diagnosis systems and reduce data processing time for clinician-computer scientist teams. Methods: A dataset of 22,132 laryngoscopic frames was extracted from 137 flexible laryngostroboscopic videos from 115 patients. 55 videos were from healthy patients with no laryngeal pathology and 82 videos were from patients with vocal fold polyps. The extracted frames were manually labeled as informative or uninformative by two independent reviewers based on vocal fold visibility, lighting, focus, and camera distance, resulting in 18,114 informative frames and 4018 uninformative frames. The dataset was split into training and test sets. A pre-trained ResNet-18 model was trained using transfer learning to classify frames as informative or uninformative. Hyperparameters were set using cross-validation. The primary outcome was precision for the informative class and secondary outcomes were precision, recall, and F1-score for all classes. The processing rate for frames between the model and a human annotator were compared. Results: The automated classifier achieved an informative frame precision, recall, and F1-score of 94.4%, 90.2%, and 92.3%, respectively, when evaluated on a hold-out test set of 4438 frames. The model processed frames 16 times faster than a human annotator. Conclusion: The CNN-based classifier demonstrates high precision for classifying informative frames in flexible laryngostroboscopic videos. This model has the potential to aid researchers with dataset creation for computer-aided diagnosis systems by automatically extracting relevant frames from laryngoscopic videos.

Applications of Artificial Intelligence to Office Laryngoscopy: A Scoping Review.

OBJECTIVES/HYPOTHESIS: This scoping review aims to provide a broad overview of the applications of artificial intelligence (AI) to office laryngoscopy to identify gaps in knowledge and guide future research. STUDY DESIGN: Scoping Review. METHODS: Searches for studies on AI and office laryngoscopy were conducted in five databases. Title and abstract and then full-text screening were performed. Primary research studies published in English of any date were included. Studies were summarized by: AI applications, targeted conditions, imaging modalities, author affiliations, and dataset characteristics. RESULTS: Studies focused on vocal fold vibration analysis (43%), lesion recognition (24%), and vocal fold movement determination (19%). The most frequently automated tasks were recognition of vocal fold nodules (19%), polyp (14%), paralysis (11%), paresis (8%), and cyst (7%). Imaging modalities included high-speed laryngeal videos (45%), stroboscopy (29%), and narrow band imaging endoscopy (7%). The body of literature was primarily authored by science, technology, engineering, and math (STEM) specialists (76%) with only 30 studies (31%) involving co-authorship by STEM specialists and otolaryngologists. Datasets were mostly from single institution (84%) and most commonly originated from Germany (23%), USA (16%), Spain (9%), Italy (8%), and China (8%). Demographic information was only reported in 39 studies (40%), with age and sex being the most commonly reported, whereas race/ethnicity and gender were not reported in any studies. CONCLUSION: More interdisciplinary collaboration between STEM and otolaryngology research teams improved demographic reporting especially of race and ethnicity to ensure broad representation, and larger and more geographically diverse datasets will be crucial to future research on AI in office laryngoscopy. LEVEL OF EVIDENCE: NA Laryngoscope, 132:1993-2016, 2022.

Times to Diagnosis, Staging, and Treatment of Head and Neck Cancer Before and During COVID-19.

OBJECTIVE: The coronavirus disease 2019 (COVID-19) pandemic has reduced the demand for, and supply of, head and neck cancer services. This study compares the times to diagnosis, staging, and treatment of head and neck cancers before and during the COVID-19 pandemic. STUDY DESIGN: Retrospective cohort study. SETTING: Tertiary academic medical center in New York City (NYC). METHODS: The times to diagnosis, staging, and treatment of head and neck cancer for patients presenting to the clinics of 4 head and neck oncology surgeons with newly diagnosed head and neck cancers were compared between pre-COVID-19 and COVID-19 periods. RESULTS: Sixty-eight patients in the pre-COVID-19 period and 26 patients in the COVID-19 period presented with newly diagnosed head and neck cancer. Patients in the COVID-19 group had a significantly longer time to diagnosis than the pre-COVID-19 group after adjustment for age and cancer diagnosis (P = .02; hazard ratio [HR], 0.54; 95% CI, 0.32-0.92). Patients in the pre-COVID-19 and COVID-19 groups had no statistically significant differences in time to staging (P > .9; HR, 1.01; 95% CI, 0.58-1.74) or time to treatment (P = .12; HR, 1.55; 95% CI, 0.89-2.72). CONCLUSION: This study found that time to diagnosis for head and neck cancers was delayed during a COVID-19 period compared to a pre-COVID-19 period. However, there was no evidence of delays in time to staging and time to treatment during the COVID-19 period. Our results prompt further investigations into the factors contributing to diagnostic delays but provide reassurance that despite COVID-19, patients were receiving timely staging and treatment for head and neck cancers.

Differences in antibiotic prescriptions between direct-to-consumer telehealth and telehealth in the emergency department.

BACKGROUND: Telemedicine, which allows physicians to assess and treat patients via real-time audiovisual conferencing, is a rapidly growing modality for providing medical care. Antibiotic stewardship is one important measure of care quality, and research on antibiotic prescribing for acute respiratory infections in direct-to-consumer telemedicine has yielded mixed results. We compared antibiotic prescription rates for acute respiratory infections in two groups treated by telemedicine: (1) patients treated via a direct-to-consumer telemedicine application and (2) patients treated via telemedicine while physically inside the emergency department. METHODS: We included direct-to-consumer telemedicine and emergency department telemedicine visits for patients 18 years and older with physician-coded International Classification of Diseases, Tenth Revision acute respiratory infection diagnoses between November 2016 and December 2018. Patients in both groups were seen by the same emergency department faculty working dedicated telemedicine shifts. We compared antibiotic prescribing rates for direct-to-consumer telemedicine and emergency department telemedicine visits before and after adjustment for age, sex, and diagnosis. RESULTS: We identified a total of 468 acute respiratory infection visits: 191 direct-to-consumer telemedicine visits and 277 emergency department telemedicine visits. Overall, antibiotics were prescribed for 47% of visits (59% of direct-to-consumer telemedicine visits vs 39% of emergency department telemedicine visits; odds ratio 2.23; 95% confidence interval 1.53-3.25; P < 0.001). The difference in antibiotic prescribing rates remained significant after adjustment for age, sex, and diagnosis (odds ratio 2.49; 95% confidence interval 1.65-3.77; P < 0.001). CONCLUSION: Patients seen by the same group of physicians for acute respiratory infection were significantly more likely to be prescribed antibiotics by direct-to-consumer telemedicine care compared with telemedicine care in the emergency department. This work suggests that contextual factors rather than evaluation over video may contribute to differences in antibiotic stewardship for direct-to-consumer telemedicine encounters.

Micro Satellite Orbital Boost by Electrodynamic Tethers.

In this manuscript, a method for maneuvering a spacecraft using electrically charged tethers is explored. The spacecraft's velocity vector can be modified by interacting with Earth's magnetic field. Through this method, a spacecraft can maintain an orbit indefinitely by reboosting without the constraint of limited propellant. The spacecraft-tether system dynamics in low Earth orbit are simulated to evaluate the effects of Lorentz force and torques on translational motion. With 500-meter tethers charged with a 1-amp current, a 100-kg spacecraft can gain 250 m of altitude in one orbit. By evaluating the combined effects of Lorenz force and the coupled effects of Lorentz torque propagation through Euler's moment equation and Newton's translational motion equations, the simulated spacecraft-tether system can orbit indefinitely at altitudes as low as 275 km. Through a rare evaluation of the nonlinear coupling of the six differential equations of motion, the one finding is that an electrodynamic tether can be used to maintain a spacecraft's orbit height indefinitely for very low Earth orbits. However, the reboost maneuver is inefficient for high inclination orbits and has high electrical power requirement. To overcome greater aerodynamic drag at lower altitudes, longer tethers with higher power draw are required.

Antibiotic Prescribing Practices: Is There a Difference Between Patients Seen by Telemedicine Versus Those Seen In-Person?

Background: Direct-to-consumer telemedicine is becoming part of mainstream medicine, but questions exist regarding the quality of care provided. We assessed antibiotic stewardship, one measure of quality, by comparing antibiotic prescription rates for acute respiratory infections (ARIs) between patients seen by telemedicine and patients seen in-person in two urban emergency departments (EDs). Methods: In two urban EDs where low-acuity patients in the ED have the option of being seen by telemedicine rather than in-person, we analyzed telemedicine and in-person visits of patients ≥18 years who received ARI diagnoses between July 2016 and September 2017. The identified ARI telemedicine visits were matched to in-person visits by diagnosis, treatment hospital, and Emergency Severity Index level. We compared antibiotic prescribing rates for telemedicine and in-person visits. Results: We identified 260 telemedicine visits and compared with 260 matched in-person visits. Antibiotics for ARIs were prescribed for 29% of telemedicine visits and 28% of in-person visits (odds ratio [OR] 1.038; 95% confidence interval [CI] 0.71-1.52; p = 0.846). This finding did not materially change after adjustment for age and gender (adjusted OR 1.034; 95% CI 0.70-1.53; p = 0.86). Conclusions: Antibiotic prescribing rates for ARIs were similar for patients seen by telemedicine and patients seen in-person at two urban EDs. If differences in antibiotic stewardship between telemedicine and in-person encounters are found, contextual factors unrelated to the video-based evaluation should be investigated.

On the mechanism of Lewis base catalyzed aldol addition reactions: kinetic and spectroscopic investigations using rapid-injection NMR.

The mechanistic foundations of the Lewis base catalyzed aldol addition reactions have been investigated. From a combination of low-temperature spectroscopic studies ((29)Si and (31)P NMR) and kinetic analyses using a rapid-injection NMR apparatus (RINMR), a correlation of the ground states and transition structures for the aldolization reactions has been formulated. The aldol addition of the tert-butylsilyl ketene acetal of tert-butyl propanoate with 1-naphthaldehyde is efficiently catalyzed by a combination of silicon tetrachloride and chiral phosphoramide Lewis bases. The rates and selectivities of the aldol additions are highly dependent on the structure of the Lewis bases: bisphosphoramides give the highest rate and selectivity, whereas a related monophosphoramide reacts slowly and with low selectivity. The monophosphoramide shows no nonlinear behavior. All of the additions show a first-order kinetic dependence on silyl ketene acetal and 1-naphthaldehyde and a zeroth-order dependence on silicon tetrachloride. The kinetic order in catalyst is structure dependent and is either half-, two-thirds-, or first-order. All of the phosphoramides are saturated with silicon tetrachloride in some form, and the resting-state species are mixtures of monomeric and dimeric, pentacoordinate cationic, or hexacoordinate neutral complexes. These data allow the formulation of a unified mechanistic scheme based on the postulate of a common reactive intermediate for all catalysts.