Artificial neural network prediction of post-thyroidectomy outcome.

OBJECTIVES: The goal of this study was to develop a deep neural network (DNN) for predicting surgical/medical complications and unplanned reoperations following thyroidectomy. DESIGN, SETTING, AND PARTICIPANTS: The 2005-2017 American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) database was queried to extract patients who underwent thyroidectomy. A DNN consisting of 10 layers was developed with an 80:20 breakdown for training and testing. MAIN OUTCOME MEASURES: Three primary outcomes of interest, including occurrence of surgical complications, medical complications, and unplanned reoperation were predicted. RESULTS: Of the 21 550 patients who underwent thyroidectomy, medical complications, surgical complications and reoperation occurred in 1723 (8.0%), 943 (4.38%) and 2448 (11.36%) patients, respectively. The DNN performed with an area under the curve of receiver operating characteristics of .783 (medical complications), .709 (surgical complications) and .703 (reoperations). Accuracy, specificity and negative predictive values of the model for all outcome variables ranged 78.2%-97.2%, while sensitivity and positive predictive values ranged 11.6%-62.5%. Variables with high permutation importance included sex, inpatient versus outpatient and American Society of Anesthesiologists class. CONCLUSIONS: We predicted surgical/medical complications and unplanned reoperation following thyroidectomy via development of a well-performing ML algorithm. We have also developed a web-based application available on mobile devices to demonstrate the predictive capacity of our models in real time.

Impact of Treatment Delay in Head and Neck Mucosal Melanoma on Overall Patient Survival.

Objectives Head and neck mucosal melanoma (HNMM) is a rare malignancy with high mortality. This study evaluates the impact of treatment delays on overall survival in HNMM. Design/Setting/Participants A retrospective review of patients with surgically managed HNMM treated with adjuvant radiation was performed from the 2004-2016 National Cancer Database. Main Outcome Measures Durations of diagnosis-to-treatment initiation (DTI), surgery-to-radiotherapy initiation (SRT), duration of radiotherapy (RTD), surgery-to-immunotherapy initiation (SIT), diagnosis-to-treatment end (DTE), and total treatment package (TTP) were calculated. Results A total of 1,011 patients (50.7% female, 90.5% Caucasian) met inclusion criteria. Median DTI, SRT, RTD, SIT, DTE, and TTP were 30, 49, 41, 102, 119, and 87 days, respectively. Only longer DTE was associated with decreased mortality (hazard ratio, 0.720; 95% confidence interval, 0.536-0.965; p = 0.028). Conclusion DTI, SRT, RTD, SIT, and TTP do not significantly affect overall survival in patients with HNMM who undergo surgery and adjuvant radiation. Longer DTE is associated with improved survival in this population. Level of Evidence 4.

Impact of esthesioneuroblastoma treatment delays on overall patient survival.

OBJECTIVES: To characterize clinical factors associated with esthesioneuroblastoma treatment delays and determine the impact of these delays on overall survival. STUDY DESIGN: Retrospective database analysis. METHODS: The 2004-2016 National Cancer Database was queried for patients with esthesioneuroblastoma managed by primary surgery and adjuvant radiation. Durations of diagnosis-to-treatment initiation (DTI), diagnosis-to-treatment end (DTE), surgery-to-RT initiation (SRT), radiotherapy treatment (RTD), and total treatment package (TTP) were analyzed. The cohort was split into two groups for each delay interval using the median time as the threshold. RESULTS: A total of 814 patients (39.6% female, 88.5% white) with mean ± SD age of 52.6 ± 15.1 years who underwent both esthesioneuroblastoma surgery and adjuvant radiotherapy were queried. Median DTI, DTE, SRT, RTD, and TTP were 34, 140, 55, 45, and 101 days, respectively. A significant association was identified between increased regional radiation dose above 66 Gy and decreased DTI (OR = 0.54, 95% CI 0.35-0.83, p = 0.01) and increased RTD (OR = 3.94, 95% CI 2.36-6.58, p < 0.001) durations. Chemotherapy administration was linked with decreased SRT (OR = 0.64, 95% CI 0.47-0.89, p = 0.01) and TTP (OR = 0.59, 95% CI 0.43-0.82, p = 0.001) durations. Cox proportional-hazards analysis revealed that increased RTD was associated with decreased survival (HR = 1.80, 95% CI 1.26-2.57, p < 0.005), independent of age, sex, race, regional radiation dose, facility volume, facility type, insurance status, modified Kadish stage, chemotherapy status, Charlson-Deyo comorbidity index, and surgical margins. CONCLUSIONS: Delays during, and prolongation of radiotherapy for esthesioneuroblastoma appears to be associated with decreased survival. LEVEL OF EVIDENCE: 4 Laryngoscope, 133:764-772, 2023.

Does Medical School Geography and Ranking Influence Residency Match in Otolaryngology?

OBJECTIVE: To elucidate the associations between geographic locations, rankings, and size/funding of medical schools and residency programs among the current otolaryngology residents. METHODS: This retrospective cross-sectional study queried otolaryngology residency program websites for relevant publicly accessible information. Location was categorized as Midwest, Northeast, South, and West. Ranking was according to Doximity (residency) and US News and World Report (medical school). Medical school and residency programs were labeled large if they had >704 students or >15 residents, respectively. RESULTS: A total of 1413 residents from 98 (89%) otolaryngology residency programs were included. Residents attending their home medical schools (18%) were equally distributed among regions (P = .845). Residents who attended medical schools in the same US regions (54%) were more likely from top-25 (P = .001) or private (P < .001) medical schools. Southern residents were most likely (64%) and Western residents were least likely (39%) from regional medical schools (P < .001), while residents from Midwest and Northeast had similar rates (54%-55%). The percentage of Midwest residents coming from regional medical schools has decreased from the 2013 to 2014 residency cycle (P = .037). Completing undergraduate school, medical school, and residency in the same region (38%) was also highest in the South (45%) and lowest in the West (25%) (P < .001). Residents at top-ranked residency programs were more likely from top-ranked (P < .001), large (P = .025), and private (P = .018) medical schools. CONCLUSION: There exist significant associations between otolaryngology residents' medical school location, ranking, size, and funding source and their residency destination. More than half of the current otolaryngology residents attended medical school in the same geographic region, and about one-fifth have attended medical school and residency at the same institution. Future studies are warranted to evaluate how these results change as the match process evolves in the future. LEVEL OF EVIDENCE: N/A.

A Comparison of Near-Infrared Imaging and Computerized Tomography Scan for Detecting Maxillary Sinusitis.

OBJECTIVE: To investigate the use of near-infrared (NIR) imaging as a tool for outpatient clinicians to quickly and accurately assess for maxillary sinusitis and to characterize its accuracy compared to computerized tomography (CT) scan. METHODS: In a prospective investigational study, NIR and CT images from 65 patients who presented to a tertiary care rhinology clinic were compared to determine the sensitivity and specificity of NIR as an imaging modality. RESULTS: The sensitivity and specificity of NIR imaging in distinguishing normal versus maxillary sinus disease was found to be 90% and 84%, normal versus mild maxillary sinus disease to be 76% and 91%, and mild versus severe maxillary sinus disease to be 96% and 81%, respectively. The average pixel intensity was also calculated and compared to the modified Lund-Mackay scores from CT scans to assess the ability of NIR imaging to stratify the severity of maxillary sinus disease. Average pixel intensity over a region of interest was significantly different (P < .001) between normal, mild, and severe disease, as well as when comparing normal versus mild (P < .001, 95% CI 42.22-105.39), normal versus severe (P < .001, 95% CI 119.43-174.14), and mild versus severe (P < .001, 95% CI 41.39-104.56) maxillary sinus disease. CONCLUSION: Based on this data, NIR shows promise as a tool for identifying patients with potential maxillary sinus disease as well as providing information on severity of disease that may guide administration of appropriate treatments.

Plastic Surgery and Music: Examining Plastic Surgery References in Hit Songs.

The purpose of this study was to assess the frequency by which plastic surgery-related terms have been included in the lyrics of Western music hits from the 1970s to the present day as a proxy for estimating the cultural impact of plastic surgery. A list of the Billboard Year-End Hot 100 songs from 1968 to 2019 and the Billboard Year-End Hot R&B/Hip-Hop songs from 1970 to 2019 was obtained for a combined total of 8550 songs. Lyrics for each song were extracted via a web-scraping system, and a database of plastic surgery-related terms was developed by our team. Each term was then queried amongst the compiled lyrics data sets, and the total frequency of plastic surgery-related terms per year and per decade was determined. Each term was also examined in its context of usage to validate its relevance to plastic surgery and determine its connotation through sentiment analysis. The frequency of plastic surgery-related terms referenced in the Billboard Year-End Hot 100 and Billboard Year-End Hot R&B/Hip-Hop charts has increased 15-fold from the 1970s (n = 1 song) to 2010s (n = 15 songs). The terms most often mentioned included "doctor," "silicone," "plastic," "surgery," "nip-tuck," and "lipo." Artists who most frequently used plastic surgery-related terms were Kanye West, 2 Chainz, and Nicki Minaj. The current study is the first to evaluate trends in plastic surgery references in music formally. In turn, this study helps further our understanding of the interplay between plastic surgery and popular culture.

Migraine Features in Patients With Isolated Aural Fullness and Proposal for a New Diagnosis.

OBJECTIVE: To compare the presence of migraine features between patients with isolated aural fullness (AF) who meet the diagnostic criteria for migraine headache and those who do not, and to propose diagnostic criteria for migraine-related AF based on our results. METHODS: We performed a retrospective study of patients presenting to a tertiary-care neurotology clinic between 2014 and 2020 with migraine-related AF. This was defined as isolated, prolonged aural fullness concurrent with migraine features once other etiologies were ruled out via examination, audiometry, and imaging. Migraine features were compared between patients meeting the diagnostic criteria for migraine headache and those not meeting the criteria. RESULTS: Seventy-seven patients with migraine-related AF were included. The mean age was 56 ±â€Š15 years and 55 (71%) patients were female. Eleven (14%) patients fulfilled the criteria for migraine headache (migraine group). Of the 66 patients who did not meet the criteria (nonmigraine group), 17 (26%) met 4/5 criteria, and 32 (48%) met 3/5 criteria, for a total of 49 (74%) patients. The migraine and nonmigraine groups were only different in 5 of 20 features, including family history of migraine (p = 0.007), sound sensitivity (p < 0.001), mental fogginess (p = 0.008), visual motion sensitivity (p = 0.008), and light sensitivity (p < 0.001). CONCLUSION: There are minimal differences in the overall prevalence of migraine features between patients with migraine-related AF who meet and do not meet the diagnostic criteria for migraine. Our findings suggest that the criteria may be too stringent and exclude many patients from potentially benefitting from treatment with migraine prophylaxis.

A Web-Based Deep Learning Model for Automated Diagnosis of Otoscopic Images.

OBJECTIVES: To develop a multiclass-classifier deep learning model and website for distinguishing tympanic membrane (TM) pathologies based on otoscopic images. METHODS: An otoscopic image database developed by utilizing publicly available online images and open databases was assessed by convolutional neural network (CNN) models including ResNet-50, Inception-V3, Inception-Resnet-V2, and MobileNetV2. Training and testing were conducted with a 75:25 breakdown. Area under the curve of receiver operating characteristics (AUC-ROC), accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were used to compare different CNN models' performances in classifying TM images. RESULTS: Our database included 400 images, organized into normal (n = 196) and abnormal classes (n = 204), including acute otitis media (n = 116), otitis externa (n = 44), chronic suppurative otitis media (n = 23), and cerumen impaction (n = 21). For binary classification between normal versus abnormal TM, the best performing model had average AUC-ROC of 0.902 (MobileNetV2), followed by 0.745 (Inception-Resnet-V2), 0.731 (ResNet-50), and 0.636 (Inception-V3). Accuracy ranged between 0.73-0.77, sensitivity 0.72-0.88, specificity 0.58-0.84, PPV 0.68-0.81, and NPV 0.73-0.83. Macro-AUC-ROC for MobileNetV2 based multiclass-classifier was 0.91, with accuracy of 66%. Binary and multiclass-classifier models based on MobileNetV2 were loaded onto a publicly accessible and user-friendly website (https://headneckml.com/tympanic). This allows the readership to upload TM images for real-time predictions using the developed algorithms. CONCLUSIONS: Novel CNN algorithms were developed with high AUC-ROCs for differentiating between various TM pathologies. This was further deployed as a proof-of-concept publicly accessible website for real-time predictions.

Automated acquisition of explainable knowledge from unannotated histopathology images.

Deep learning algorithms have been successfully used in medical image classification. In the next stage, the technology of acquiring explainable knowledge from medical images is highly desired. Here we show that deep learning algorithm enables automated acquisition of explainable features from diagnostic annotation-free histopathology images. We compare the prediction accuracy of prostate cancer recurrence using our algorithm-generated features with that of diagnosis by expert pathologists using established criteria on 13,188 whole-mount pathology images consisting of over 86 billion image patches. Our method not only reveals findings established by humans but also features that have not been recognized, showing higher accuracy than human in prognostic prediction. Combining both our algorithm-generated features and human-established criteria predicts the recurrence more accurately than using either method alone. We confirm robustness of our method using external validation datasets including 2276 pathology images. This study opens up fields of machine learning analysis for discovering uncharted knowledge.

Illuminating Clues of Cancer Buried in Prostate MR Image: Deep Learning and Expert Approaches.

Deep learning algorithms have achieved great success in cancer image classification. However, it is imperative to understand the differences between the deep learning and human approaches. Using an explainable model, we aimed to compare the deep learning-focused regions of magnetic resonance (MR) images with cancerous locations identified by radiologists and pathologists. First, 307 prostate MR images were classified using a well-established deep neural network without locational information of cancers. Subsequently, we assessed whether the deep learning-focused regions overlapped the radiologist-identified targets. Furthermore, pathologists provided histopathological diagnoses on 896 pathological images, and we compared the deep learning-focused regions with the genuine cancer locations through 3D reconstruction of pathological images. The area under the curve (AUC) for MR images classification was sufficiently high (AUC = 0.90, 95% confidence interval 0.87-0.94). Deep learning-focused regions overlapped radiologist-identified targets by 70.5% and pathologist-identified cancer locations by 72.1%. Lymphocyte aggregation and dilated prostatic ducts were observed in non-cancerous regions focused by deep learning. Deep learning algorithms can achieve highly accurate image classification without necessarily identifying radiological targets or cancer locations. Deep learning may find clues that can help a clinical diagnosis even if the cancer is not visible.

Stress-Induced Cellular Clearance Is Mediated by the SNARE Protein ykt6 and Disrupted by α-Synuclein.

Age-related neurodegenerative disorders are characterized by a slow, persistent accumulation of aggregated proteins. Although cells can elicit physiological responses to enhance cellular clearance and counteract accumulation, it is unclear how pathogenic proteins evade this process in disease. We find that Parkinson's disease α-synuclein perturbs the physiological response to lysosomal stress by impeding the SNARE protein ykt6. Cytosolic ykt6 is normally autoinhibited by a unique farnesyl-mediated regulatory mechanism; however, during lysosomal stress, it activates and redistributes into membranes to preferentially promote hydrolase trafficking and enhance cellular clearance. α-Synuclein aberrantly binds and deactivates ykt6 in patient-derived neurons, thereby disabling the lysosomal stress response and facilitating protein accumulation. Activating ykt6 by small-molecule farnesyltransferase inhibitors restores lysosomal activity and reduces α-synuclein in patient-derived neurons and mice. Our findings indicate that α-synuclein creates a permissive environment for aggregate persistence by inhibiting regulated cellular clearance and provide a therapeutic strategy to restore protein homeostasis by harnessing SNARE activity.