Background: Despite the rapid evolvement of machine learning (ML) applications within the medical literature, there exist a paucity of knowledge regarding diagnosis, decision-making, and predicting surgical outcomes in otolaryngology using ML models. This thesis aims to 1) Construct ML models that use pre-operative-only inputs to predict length of stay (LOS) and discharge to non-home facility (DNHF) following complex head and neck (HN) surgeries, 2) Utilize deep learning and a large number of input variables to predict short-term adverse events following vestibular schwannoma (VS) surgery, 3) Construct preliminary binary (normal vs. abnormal) and multiclass (normal vs. acute otitis media vs. otitis externa vs. chronic suppurative otitis media vs. cerumen) image classification models of otoscopic images, and 4) Publish the best-performing models as public web-based interfaces.

Methods: To develop novel ML models that predict various post-operative outcomes, the 2005-2017 National Surgical Quality Improvement Program database was utilized to extract subjects undergoing major HN surgery (N=2667) and VS surgery (N=1783). Datasets were randomly stratified into training and testing sets using an 80:20 ratio with k-fold cross-validation used for training. To develop the novel otoscopic image classification model utilizing Inception-Resnet-V2 networks, 400 publicly available and labeled otoscopic images were obtained.

Results: Four classification models for predicting DNHF were developed with high specificities (range 0.80-0.84), where the generalized linear model and gradient boosting machine models outperformed artificial neural network and random forest models with receiver operating characteristic (ROC), accuracy, and negative predictive value (NPV) of 0.72-0.73, 0.75-0.76, and 0.88-0.89, respectively. Four regression models were developed for predicting LOS in days, where all performed similarly with mean absolute error and root mean squared errors of 3.95-3.98 and 5.14-5.16, respectively. The DNHF and LOS models were developed into a web-based interface: https://uci-ent.shinyapps.io/head-neck/. Using pre-, peri-, and post-operative inputs, three deep-learning models to predict unplanned reoperation, surgical complications, and medical complications following VS surgery were developed with ROC of 0.74, 0.70, and 0.83, and accuracy of 0.91, 0.71, and 0.92, respectively. Lastly, a preliminary image classification model with ROC, accuracy, and NPV of 0.89, 0.77, and 0.74 for binary classification, and the capability to suggest the diagnosis among abnormal images, was developed: https://headneckml.com/tm/.

Conclusion: Novel ML models to predict DNHF or LOS following complex HN surgery, reoperation or complications following VS surgery, and abnormalities on otoscopic imaging were successfully developed. Publishing such models as interactive web-based interfaces will help advance this frontier by providing opportunities for examination/validation and practical benefit to clinicians.

Background: For the first time in 2013, the international classification of headache disorders (ICHD) published a diagnostic criterion for vestibular migraine (VM). This provided, legitimacy to a diagnosis previously rejected by many, further driving a discussion regarding its validity.

Objective: To characterize a cohort of patients suspected of having VM.

Methods: Spanning a 3-year period, questionnaires were provided to patients at an outpatient neurotology clinic assessing ICHD diagnostic criteria for VM, and migraine.

Results: Of the 396 participants included, 26% met criteria for VM, 25% met criteria for migraine only with no VM (MO), and 48% met no criteria (MNC) for either migraine or VM. Significant overlap in vertigo symptoms was observed in the MO group compared to VM participants, with only 3 of 11 features proving to be statistically different between the two. Similarly, 21 of 23 features of migraine were equivalent between the MO and VM group. In contrast, MNC and VM groups were significantly different in almost all features relating to migraine, vertigo and quality of life.

Conclusions: Evidence of the significant symptomatic overlap observed in VM and MO groups demonstrates the two could be suffering from an identical disease process. In addition, the MNC group may belong to a benign recurrent vertigo entity due to a lack of features necessary to achieve ICHD criteria for migraine or VM. In all cases, there is evidence that migraine may play a significant role in vertigo symptoms, leading to a possible therapeutic improvement with migraine prophylaxis.

Purpose of review

Recent findings

Summary

Neuroprosthetic devices have become increasingly important in the treatment of clinical pathologies. In particular, the application of neuroprosthetic devices to restore function following damage to peripheral and cranial nerves has been demonstrated to be efficacious, safe, and has been readily adopted in the clinic to treat a wide range of pathologies. Neuroprosthetic devices have been successfully deployed in numerous cranial nerves, including the hypoglossal nerve to treat obstructive sleep apnea and the vagus nerve to treat epilepsy, heart failure, and manage blood pressure. However, the application of neuroprosthetic devices to the facial nerve (FN) for purposes of facial reanimation has not been explored.

FN injury can cause debilitating and permanent damage with oftentimes limited treatment options. In cases of FN injury, patients may develop permanent facial paralysis (FP), which can cause facial muscles to lose tone, and over time, atrophy and convert into scar tissue. FP arises from a variety of causes, including tumor, surgery, trauma or infection, and results in problems with eye irritation, visual impairment, drooling, intraoral food retention, and demoralizing cosmetic deformities.

In recent decades, considerable efforts have been undertaken to care for patients with permanent FP. Several surgical interventions, including static and dynamic options, have been described for patients with unilateral FP. However, these interventions only address specific parts of the face, have a 10-15% failure rate, and require multiple procedures involving multiple surgical sites to accomplish functional goals. An alternative approach to facial reanimation is the utilization of neuroprosthetic technologies to deliver exogenous current to the partially recovered FN in order to provide both facial tone and selective activation of individual muscles.

We demonstrate the ability of both multichannel microelectrode arrays and multichannel cuff electrodes to selectively stimulate the feline FN in both acute and chronic settings, and couple this technology with closed loop circuitry to develop a putative approach to volitional and graded hemifacial animation. We also present data from clinical investigations on human FN selectivity, and investigate the application of these technologies in the recurrent laryngeal nerve.