Deep Learning Model for Real­time Semantic Segmentation During Intraoperative Robotic Prostatectomy.

Background and objective: Recently, deep learning algorithms, including convolutional neural networks (CNNs), have shown remarkable progress in medical imaging analysis. Semantic segmentation, which segments an unknown image into different parts and objects, has potential applications in robotic surgery in areas where artificial intelligence (AI) can be applied, such as in AI-assisted surgery, surgeon training, and skill assessment. We aimed to investigate the performance of a CNN-based deep learning model in real-time segmentation in robot-assisted radical prostatectomy (RALP). Methods: Intraoperative videos of RALP procedures were obtained. The reinforcement U-Net model was used for segmentation. Segmentation of the images of instruments, bladder, prostate, and seminal vesicle-vas deferens was performed. The dataset was preprocessed and split randomly into training, validation, and test data in a 7:2:1 ratio. Dice coefficient, intersection over union (IoU), and accuracy by class, which are commonly used in medical image segmentation, were calculated to evaluate the performance of the model. Key findings and limitations: From 120 patient videos, 2400 images were selected for RALP procedures. The mean Dice scores for the identification of the instruments, bladder, prostate, and seminal vesicle-vas deferens were 0.96, 0.74, 0.85, and 0.84, respectively. Overall, when applied to the test data, the model had a mean Dice coefficient value of 0.85, IoU of 0.77, and accuracy of 0.85. Limitations included the sample size, lack of diversity in the methods of surgery, incomplete surgical processes, and lack of external validation. Conclusions and clinical implications: The CNN-based segmentation provides accurate real-time recognition of surgical instruments and anatomy in RALP. Deep learning algorithms can be used to identify anatomy within the surgical field and could potentially be used to provide real-time guidance in robotic surgery. Patient summary: We demonstrate the potential effectiveness of deep learning segmentation in robotic prostatectomy procedures. Deep learning algorithms could be used to identify anatomical structures within the surgical field and may provide real-time guidance in robotic surgery.

Novel Molecular Genetic Etiology of Asymmetric Hearing Loss: Autosomal-Dominant LMX1A Variants.

INTRODUCTION: Sensorineural hearing loss is the most common sensory disorder in humans. Genetic analyses have greatly increased our understanding of the pathogenic mechanisms in play. Thus, characterization of audiologic phenotypes by the genetic etiology may aid elucidation of the etiologies of certain types of inherited hearing loss. Further, delineation of specific audiologic phenotypes based on the genetic etiology aids our understanding of some types of inherited hearing loss in terms of the prediction of clinical course, revelation of genotype-phenotype correlations, and application of appropriate audiologic rehabilitation. Here, we describe the interesting audiologic characteristics of LMX1A -associated deafness, which revealed significant asymmetry between two ears. METHODS: Among 728 probands of which genomic DNA went through exome sequencing regardless of any specific audiologic phenotypes, probands for which exome sequencing was performed and a causative LMX1A variant was found were all included. Five LMX1A -associated DFNA7 families (approximately 0.7%), the pedigrees of whom indicated autosomal-dominant hearing loss, were identified, and segregation was studied using Sanger sequencing. The affected individuals underwent comprehensive evaluations, including medical history reviews, physical examinations, imaging, and auditory phenotyping. We functionally characterized the novel LMX1A variants via computational structural modeling and luciferase reporter assays. RESULTS: Among 728 probands of which genomic DNA went through exome sequencing, we identified four novel LMX1A heterozygous variants related to DFNA7 (c.622C>T:p.Arg208*, c.719A>G:p.Gln240Arg, c.721G>A:p.Val241Met, and c.887dup:p.Gln297Thrfs*41) and one harboring a de novo heterozygous missense LMX1A variant (c.595A>G;p.Arg199Gly) previously reported. It is important to note that asymmetric hearing loss was identified in all probands and most affected individuals, although the extent of asymmetry varied. Structural modeling revealed that the two missense variants, p.Gln240Arg and p.Val241Met, affected conserved residues of the homeodomain, thus attenuating LMX1A-DNA interaction. In addition, Arg208*-induced premature termination of translation destroyed the structure of the LMX1A protein, including the DNA-binding homeodomain, and p.Gln297Thrfs*41 led to the loss of the C-terminal helix involved in LIM2 domain interaction. Compared with the wild-type protein, all mutant LMX1A proteins had significantly reduced transactivation efficiency, indicating that the ability to elicit transcription of the downstream target genes of LMX1A was severely compromised. Thus, in line with the American College of Medical Genetics and Genomics guideline specified to genetic hearing loss, the four novel LMX1A variants were identified as "pathogenic" (p.Arg208* and p.Gln297Thrfs*41), "likely pathogenic" (p.Val241Met), and as a "variant of uncertain significance'' (p.Gln240Arg). CONCLUSION: For the first time, we suggest that LMX1A is one of the candidate genes which, if altered, could be associated with dominantly inherited asymmetric hearing loss. We also expand the genotypic spectrum of disease-causing variants of LMX1A causing DFNA7 by doubling the number of LMX1A variants reported thus far in the literature.

Modiolar Proximity of Slim Modiolar Electrodes and Cochlear Duct Length: Correlation for Potential Basis of Customized Cochlear Implantation With Perimodiolar Electrodes.

OBJECTIVES: Recent studies have shown that cochlear duct length (CDL) varies among individuals and could significantly influence the final position of the electrode and its trajectory in the cochlea. Given this, we hypothesized that the degree of modiolar proximity of novel slim modiolar electrodes, such as CI532 and CI632, can also be affected by CDL. To test this hypothesis, we retrospectively evaluated individual CDL to determine if there is any significant correlation of CDL with degree of modiolar proximity. METHODS: Fifty-one ears from 38 subjects implanted with slim modiolar electrodes by a single surgeon through the round window approach using the pull-back technique were included. Our cohort was classified according to the deafness onset (congenital versus postlingual) and the degree of modiolar proximity (less versus tight) with reference to the spiral diameter made by the slim modiolar electrodes in situ on transorbital x ray. We then analyzed the CDL and its metrics using a readily available surgical preplanning tool (OTOPLAN) to obtain comparable data. RESULTS: Among 30 ears associated with congenital deafness, 9 ears (30%) showed less modiolar proximity, while none of the 21 ears from 19 subjects with postlingual deafness exhibited "less modiolar proximity" based on our criteria. In this study, CDL showed significant variation among subjects. Importantly, a significant inverse correlation between spiral diameter and CDL (ρ = -0.581, p < 0.001) was found, showing that shorter CDLs have longer spiral diameter and less modiolar proximity. Moreover, further pull-back technique characterized by pulling out the electrode a little bit more in cases with shorter CDL, if not always, exhibited tighter modiolar proximity. CONCLUSION: A preponderance of less modiolar proximity of the electrode was observed exclusively among congenital deafness cases, demonstrated by a less tight spiral configuration even under the pull-back technique. Our data suggest that shorter CDL is associated with a less tight spiral configuration of slim modiolar electrodes postoperatively. Depending on the insertion technique, the differential degree of modiolar proximity of slim modiolar electrodes can be alleviated in cases with short CDL, which justifies cochlear duct length-based customized insertion of slim modiolar electrodes.

Anxiety symptoms and occupational stress among young Korean female manufacturing workers.

BACKGROUND: The prevalence of anxiety disorders has been increasing in South Korea, with recent studies reporting anxiety disorders as the most common mental disorder among all South Korean females. Anxiety disorders, which are independent risk factors of suicidal ideation and suicide attempts, are significantly correlated with productivity loss, high medical costs, impaired work performance, and frequent worker absence, and thus are potentially serious problems affecting the health of South Korean female workers. In previous studies, anxiety disorders were shown to have a significant correlation with occupational stress. This study seeks to examine the prevalence of anxiety symptoms as well as the relationship between occupational stress and anxiety symptoms among South Korean female manufacturing workers. METHODS: A structured self-reported questionnaire was administered to 1,141 female workers at an electrical appliance manufacturing plant. The questionnaire collected data on general characteristics, health behaviors, sleep quality, job characteristics (shift work, shift work schedule, and job tenure), occupational stress, and anxiety symptoms. Sleep quality was measured using the Pittsburgh Sleep Quality Index, occupational stress with the Korean Occupational Stress Scale-Short Form (KOSS-SF), and anxiety symptoms with the Korean version of the Beck Anxiety Inventory. A chi square test was conducted to determine the distribution differences in anxiety symptoms based on general characteristics, health behaviors, job characteristics, and sleep quality. A linear-by-linear association test was used to determine the distribution differences between anxietysymptoms and the levels of occupational stress. Last, logistic regression analysis was used in order to determine the association between occupational stress and anxiety symptoms. RESULTS: The prevalence of anxiety symptoms was 15.2 %. In the multivariate logistic regression analysis that adjusted for sleep quality and general characteristics, a significantassociation was found for those with anxiety disorders; the odds ratios (OR) were significantly higher the greater the total KOSS-SF score (moderate-risk group OR=2.85, 95 % CI=1.79-4.56; high-risk group OR=5.34, 95 % CI=3.59-7.96). In addition, excluding insufficient job control, all other KOSS-SF subscales were significantly associated with anxiety symptoms, and a relatively high OR was seen in the high-risk group for job demand (OR=3.19, 95 % CI=2.27-4.49), job insecurity (OR=4.52, 95 % CI=2.86-7.13), and occupational culture (OR=4.52, 95 % CI=2.90-7.04). CONCLUSION: There was a significant association between anxiety symptoms and occupational stress stemming from the psychosocial work environment among these South Korean female manufacturing workers. Future longitudinal studies are needed to examine the association between the occupational stress caused by the psychosocial work environment and the incidence of anxiety disorders and anxiety symptoms. Furthermore, intervention programs that aim to address the prevalence of anxiety symptoms and improve the psychosocial work environment, especially for younger female manufacturing workers, are needed.

Feasibility of optic nerve sheath diameter measured on initial brain computed tomography as an early neurologic outcome predictor after cardiac arrest.

OBJECTIVES: Few parameters are available to predict neurologic outcome of post-cardiac arrest patients in the early stage of treatment. Optic nerve sheath diameter (ONSD) has been used to indirectly assess intracranial pressure. This study evaluated whether ONSD, an additional parameter in initial brain computed tomography (CT) scans, can be an early predictor of neurologic outcome in post-cardiac arrest patients. METHODS: A total of 112 cardiac arrest patients between November 2012 and October 2013 were identified. Ninety-eight comatose cardiac arrest patients were evaluated with brain CT. Of these patients, after exclusion of patients whose brain CT scans were done too late or with poor baseline neurology (Cerebral Performance Category [CPC] ≥ 3), 91 patients were included for this study. The parameters of initial brain CT, i.e., gray matter-to-white matter ratio (GWR) and ONSD, were measured after clinical care as part of a retrospective reanalysis of images. ONSD on brain CT was bilaterally measured 3 mm behind the eyeball at fixed window width and level and averaged to yield the mean value. The performance of ONSD to predict poor neurologic outcome (CPC = 3 to 5) was analyzed using multiple logistic regression analysis, receiver operating characteristic (ROC) curve analysis, and cross-tabulations. RESULTS: Twenty-three patients showed good neurologic outcomes at hospital discharge. Mean (±SD) ONSD was 5.6 (±0.3) mm in the good outcome group versus 6.3 (±0.5) mm in the poor outcome group (p < 0.001). After basic clinical covariates were controlled for, i.e., age, Glasgow Coma Scale (GCS) score (3 vs. 4-15), and time from collapse to return of spontaneous circulation (ROSC), ONSD (odds ratio [OR] = 2.1; 95% confidence interval [CI] = 1.1 to 3.9) and GWR (OR = 0.6; 95% CI = 0.4 to 0.9) were found to be significant factors for predicting poor neurologic outcome. ROC curve analysis showed that ONSD and GWR had areas under the ROC curve of 0.931 (95% CI = 0.87 to 0.98) and 0.922 (95% CI = 0.86 to 0.97), respectively. Combining the cutoff values of ONSD (6.21 mm, sensitivity = 56%; 95% CI = 43% to 68%) and GWR (1.23, sensitivity = 84%; 95% CI = 73% to 92%) to have 100% specificities, the sensitivity was improved to 92% (95% CI = 84% to 98%). Intrarater and interrater intraclass correlation coefficients between the investigators measuring ONSD were 0.888 and 0.833, respectively. CONCLUSIONS: Optic nerve sheath diameter on initial brain CT correlated closely with the neurologic outcome of hypoxic ischemic encephalopathy and had good reliability. Additional prospective work may be justified evaluating the standardization and diagnostic performance in real time use as a predictive tool for neurologic outcome following cardiac arrest.