aEYE: A deep learning system for video nystagmus detection.

Background: Nystagmus identification and interpretation is challenging for non-experts who lack specific training in neuro-ophthalmology or neuro-otology. This challenge is magnified when the task is performed via telemedicine. Deep learning models have not been heavily studied in video-based eye movement detection. Methods: We developed, trained, and validated a deep-learning system (aEYE) to classify video recordings as normal or bearing at least two consecutive beats of nystagmus. The videos were retrospectively collected from a subset of the monocular (right eye) video-oculography (VOG) recording used in the Acute Video-oculography for Vertigo in Emergency Rooms for Rapid Triage (AVERT) clinical trial (#NCT02483429). Our model was derived from a preliminary dataset representing about 10% of the total AVERT videos (n = 435). The videos were trimmed into 10-sec clips sampled at 60 Hz with a resolution of 240 × 320 pixels. We then created 8 variations of the videos by altering the sampling rates (i.e., 30 Hz and 15 Hz) and image resolution (i.e., 60 × 80 pixels and 15 × 20 pixels). The dataset was labeled as "nystagmus" or "no nystagmus" by one expert provider. We then used a filtered image-based motion classification approach to develop aEYE. The model's performance at detecting nystagmus was calculated by using the area under the receiver-operating characteristic curve (AUROC), sensitivity, specificity, and accuracy. Results: An ensemble between the ResNet-soft voting and the VGG-hard voting models had the best performing metrics. The AUROC, sensitivity, specificity, and accuracy were 0.86, 88.4, 74.2, and 82.7%, respectively. Our validated folds had an average AUROC, sensitivity, specificity, and accuracy of 0.86, 80.3, 80.9, and 80.4%, respectively. Models created from the compressed videos decreased in accuracy as image sampling rate decreased from 60 Hz to 15 Hz. There was only minimal change in the accuracy of nystagmus detection when decreasing image resolution and keeping sampling rate constant. Conclusion: Deep learning is useful in detecting nystagmus in 60 Hz video recordings as well as videos with lower image resolutions and sampling rates, making it a potentially useful tool to aid future automated eye-movement enabled neurologic diagnosis.

Artificial Intelligence Educational & Research Initiatives and Leadership Positions in Academic Radiology Departments.

PURPOSE: Rapid advances in artificial intelligence (AI) have generated significant interest in the radiology community. However, formal AI initiatives and leadership roles in academic radiology has not been formally evaluated. The purpose of this study was to assess the prevalence of formal AI initiatives and leadership roles in academic radiology departments. METHODS: Radiology departments with National Institutes of Health funding in fiscal year 2019 were identified. AI educational and research initiatives, leadership roles, and industry partnerships were assessed by searching department websites for AI-related keywords. Correlations between NIH funding and the presence of AI initiatives were evaluated with linear regression. RESULTS: Sixty-two radiology departments with NIH funding were included in this study. Educational initiatives on AI were offered by 29 (47%) departments. Fifty-five (89%) departments had at least 1 AI researcher and 34 (55%) departments were affiliated with an AI research laboratory, center, or cluster. AI-specific leadership roles and industry partnerships were identified in 3 (5%) and 23 (37%) departments, respectively. The amount of NIH funding did not have a significant linear correlation with educational initiatives (P = 0.08) but there was a significant linear correlation between funding and presence of research initiatives (P = 0.003). CONCLUSIONS: AI educational initiatives were offered by almost half of radiology departments. Most departments had AI researchers and affiliated labs, but the majority were not led by the radiology department, and few had formal AI leadership roles. In the new AI era, these findings provide a benchmark for departments considering implementing formal AI initiatives.

Implementation science for the adductor canal block: A new and adaptable methodology process.

BACKGROUND: Following the successful Perioperative Surgical Home (PSH) practice for total knee arthroplasty (TKA) at our institution, the need for continuous improvement was realized, including the deimplementation of antiquated PSH elements and introduction of new practices. AIM: To investigate the transition from femoral nerve blocks (FNB) to adductor canal nerve blocks (ACB) during TKA. METHODS: Our 13-month study from June 2016 to 2017 was divided into four periods: a three-month baseline (103 patients), a one-month pilot (47 patients), a three-month implementation and hardwiring period (100 patients), and a six-month evaluation period (185 patients). In total, 435 subjects were reviewed. Data within 30 postoperative days were extracted from electronic medical records, such as physical therapy results and administration of oral morphine equivalents (OME). RESULTS: Our institution reduced FNB application (64% to 3%) and increased ACB utilization (36% to 97%) at 10 mo. Patients in the ACB group were found to have increased ambulation on the day of surgery (4.1 vs 2.0 m) and lower incidence of falls (0 vs 1%) and buckling (5% vs 27%) compared with FNB patients (P < 0.05). While ACB patients (13.9) reported lower OME than FNB patients (15.9), the difference (P = 0.087) did not fall below our designated statistical threshold of P value < 0.05. CONCLUSION: By demonstrating closure of the "knowledge to action gap" within 6 mo, our institution's findings demonstrate evidence in the value of implementation science. Physician education, technical support, and performance monitoring were deemed key facilitators of our program's success. Expanded patient populations and additional orthopedic procedures are recommended for future study.

Percutaneous Cryoablation for Stage 1 Renal Cell Carcinoma: Outcomes from a 10-year Prospective Study and Comparison with Matched Cohorts from the National Cancer Database.

Background Percutaneous cryoablation (PCA) is an increasingly utilized treatment for stage I renal cell carcinoma (RCC), albeit without supportive level I evidence. Purpose Primary objective was to determine the 10-year oncologic outcomes of PCA for stage I RCC in a prospective manner. Secondary objectives were to compare outcomes after partial nephrectomy (PN) and radical nephrectomy (RN) from the National Cancer Database (NCDB), to determine long-term renal function, and to determine the risk of metachronous disease. Materials and Methods In this institutional review board-approved prospective observational study (2006-2013), study participants with single, sporadic, biopsy-proven RCC were included to calculate the 10-year overall survival, recurrence-free survival, and disease-specific survival after PCA. Results were compared with matched PN and RN NCDB cohorts. Overall and recurrence-free survival probabilities were estimated by using nonparametric maximum likelihood estimator. Disease-specific survival was estimated by using the redistribution-to-right method. Age at diagnosis was stratified as a risk for survival. The effect on estimated glomerular filtration rate, serum creatinine level, and the risk for hemodialysis and metachronous disease were calculated. Results One hundred thirty-four patients (46% men) with single, sporadic, biopsy-proven RCC (median size ± standard deviation, 2.8 cm ± 1.4) were included. Overall survival was 86% (95% confidence interval [CI]: 80%, 93%) and 72% (95% CI: 62%, 83%), recurrence-free survival was 85% (95% CI: 79%, 91%) and 69% (95% CI: 59%, 79%) (improved over surgery), and disease-specific survival was 94% (95% CI: 90%, 98%) at both 5 years and 10 years (similar to surgery), respectively. The 10-year risk of hemodialysis was 2.3%. Risk of metachronous RCC was 6%. Charlson/Deyo Combined Comorbidity score analysis showed decreasing overall survival with increasing comorbidity index. The PCA cohort outperformed both RN- and PN-matched subgroups in all Charlson/Deyo Combined Comorbidity score categories. Conclusion Percutaneous cryoablation yielded a 10-year disease-specific survival of 94%, equivalent to that reported after radical or partial nephrectomy. Overall survival probability after percutaneous cryoablation at 5 years and 10 years was longer than for radical or partial nephrectomy, especially for patients at higher risk (Charlson/Deyo Combined Comorbidity score ≥2). © RSNA, 2020.