Automated detection of steps in videos of strabismus surgery using deep learning.

BACKGROUND: Learning to perform strabismus surgery is an essential aspect of ophthalmologists' surgical training. Automated classification strategy for surgical steps can improve the effectiveness of training curricula and the efficient evaluation of residents' performance. To this end, we aimed to develop and validate a deep learning (DL) model for automated detecting strabismus surgery steps in the videos. METHODS: In this study, we gathered 479 strabismus surgery videos from Shanghai Children's Hospital, affiliated to Shanghai Jiao Tong University School of Medicine, spanning July 2017 to October 2021. The videos were manually cut into 3345 clips of the eight strabismus surgical steps based on the International Council of Ophthalmology's Ophthalmology Surgical Competency Assessment Rubrics (ICO-OSCAR: strabismus). The videos dataset was randomly split by eye-level into a training (60%), validation (20%) and testing dataset (20%). We evaluated two hybrid DL algorithms: a Recurrent Neural Network (RNN) based and a Transformer-based model. The evaluation metrics included: accuracy, area under the receiver operating characteristic curve, precision, recall and F1-score. RESULTS: DL models identified the steps in video clips of strabismus surgery achieved macro-average AUC of 1.00 (95% CI 1.00-1.00) with Transformer-based model and 0.98 (95% CI 0.97-1.00) with RNN-based model, respectively. The Transformer-based model yielded a higher accuracy compared with RNN-based models (0.96 vs. 0.83, p < 0.001). In detecting different steps of strabismus surgery, the predictive ability of the Transformer-based model was better than that of the RNN. Precision ranged between 0.90 and 1 for the Transformer-based model and 0.75 to 0.94 for the RNN-based model. The f1-score ranged between 0.93 and 1 for the Transformer-based model and 0.78 to 0.92 for the RNN-based model. CONCLUSION: The DL models can automate identify video steps of strabismus surgery with high accuracy and Transformer-based algorithms show excellent performance when modeling spatiotemporal features of video frames.

What can we learn from a Chinese social media used by glaucoma patients?

PURPOSE: Our study aims to discuss glaucoma patients' needs and Internet habits using big data analysis and Natural Language Processing (NLP) based on deep learning (DL). METHODS: In this retrospective study, we used web crawler technology to crawl glaucoma-related topic posts from the glaucoma bar of Baidu Tieba, China. According to the contents of topic posts, we classified them into posts with seeking medical advice and without seeking medical advice (social support, expressing emotions, sharing knowledge, and others). Word Cloud and frequency statistics were used to analyze the contents and visualize the keywords of topic posts. Two DL models, Bidirectional Long Short-Term Memory (Bi-LSTM) and Bidirectional Encoder Representations from Transformers (BERT), were trained to identify the posts seeking medical advice. The evaluation matrices included: accuracy, F1 value, and the area under the ROC curve (AUC). RESULTS: A total of 10,892 topic posts were included, among them, most were seeking medical advice (N = 7071, 64.91%), and seeking advice regarding symptoms or examination (N = 4913, 45.11%) dominated the majority. The following were searching for social support (N = 2362, 21.69%), expressing emotions (N = 497, 4.56%), and sharing knowledge (N = 527, 4.84%) in sequence. The word cloud analysis results showed that ocular pressure, visual field, examination, and operation were the most frequent words. The accuracy, F1 score, and AUC were 0.891, 0.891, and 0.931 for the BERT model, 0.82, 0.821, and 0.890 for the Bi-LSTM model. CONCLUSION: Social media can help enhance the patient-doctor relationship by providing patients' concerns and cognition about glaucoma in China. NLP can be a powerful tool to reflect patients' focus on diseases. DL models performed well in classifying Chinese medical-related texts, which could play an important role in public health monitoring.

Comparison of iridolenticular contact area under different light brightness between post primary acute angle closure and cataract eyes.

BACKGROUND: To study the iridolenticular contact area (ILCA) under different light conditions in acute primary angle closure (APAC). METHODS: This cross-sectional, observational study involved 22 unilateral APAC patients and 59 cataract patients (59 eyes). Images of the APAC eyes, fellow eyes and cataract eyes were collected by anterior segment optical coherence tomography (ASOCT) under different light conditions respectively. The ILCA, anterior chamber width (ACW), anterior chamber area (ACA), lens vault (LV), angle opening distance at 750 µm (AOD750), trabecular iris space area at 750 µm (TISA750) and iris area at 750 µm (IA750) were measured using Image J software. RESULTS: The ILCA of cataract eyes were significantly larger than APAC eyes (4.424 ± 1.208 vs 4.049 ± 2.725mm2, P = 0.034) and fellow eyes (4.424 ± 1.208 vs 3.651 ± 1.629 mm2, P = 0.008) under dark condition. Under dark condition, ILCA of APAC eyes was negatively correlated with AOD750 (r = -0.444, P = 0.038), TISA750 (r = -0.498, P = 0.018). The ILCA of cataract eyes under dark condition was significantly greater than under bright condition (4.424 ± 1.208 vs 2.526 ± 0.992 mm2, P < 0.001). CONCLUSIONS: This study showed that ILCA in both APAC eye and fellow eye were smaller than cataract eye. Future study should focus on both the contact area and force at the interface of lens and iris with larger sample size.

Assessing the efficacy of myopia control in monocular orthokeratology treated unilateral myopic children.

PURPOSE: To investigate the efficacy of myopia control by comparing the orthokeratology (Ortho-K) treated eyes and the emmetropic contralateral eyes in unilateral myopic children, and to identify the inter-individual influence factors. METHOD: In this retrospective study, 1566 medical records of children wearing Ortho-K lens were reviewed, and 62 children who received monocular Ortho-K lens for more than 1 year were analyzed. The change in axial length (AL) of the Ortho-K eyes and the emmetropic contralateral eyes was recorded. To evaluate the absolute and relative efficacy of myopia control, the intra-bilateral absolute reduction in AL growth (ibARAL) and the intra-bilateral relative reduction in AL growth (ibRRAL) were calculated as main outcomes. Association of the AL elongation, ibARAL and ibRRAL with age, sex and ocular parameters was analyzed by correlation analysis and generalized estimating equation (GEE) analysis. RESULT: The average initial wearing age was 10.76 ± 1.45 (ranged 8.5 to 15.8). The average baseline SER was - 2.15 ± 1.03 (ranged - 5.25 to -1.00) D in the Ortho-K eyes and - 0.01 ± 0.40 (ranged - 0.75 to 0.75) D in the contralateral eyes. At the 1-year follow-up, the average increased AL was significantly less in the Ortho-K eyes (0.07 ± 0.18 mm) than in the fellow eyes (0.48 ± 0.24 mm) (p < 0.001). The mean ibARAL was 0.41 ± 0.30 mm, and the mean ibRRAL was 83.4%±56.3%. In the GEE model, the AL change in Ortho-K eyes (ß = 0.051, p = 0.009, 95%CI: 0.012 to 0.090), the ibARAL (ß= -0.153, p = 0.000, 95%CI: -0.228 to -0.078) and the ibRRAL (ß= -0.196, p = 0.020, 95%CI: -0.361 to -0.030) were independently associated with the spherical equivalent refraction (SER) of the Ortho-K eyes, after adjusting for age, sex, and keratometry. CONCLUSION: In our study, the Ortho-K treatment was efficacious in controlling axial length growth in the monocular orthokeratology treated unilateral myopic eyes. The efficacy increased when the myopia was more severe. In the children from 8 to 16 years old, the effectiveness was independent of age and sex.

Detection of shallow anterior chamber depth from two-dimensional anterior segment photographs using deep learning.

BACKGROUND: The purpose of this study was to implement and evaluate a deep learning (DL) approach for automatically detecting shallow anterior chamber depth (ACD) from two-dimensional (2D) overview anterior segment photographs. METHODS: We trained a DL model using a dataset of anterior segment photographs collected from Shanghai Aier Eye Hospital from June 2018 to December 2019. A Pentacam HR system was used to capture a 2D overview eye image and measure the ACD. Shallow ACD was defined as ACD less than 2.4 mm. The DL model was evaluated by a five-fold cross-validation test in a hold-out testing dataset. We also evaluated the DL model by testing it against two glaucoma specialists. The performance of the DL model was calculated by metrics, including accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). RESULTS: A total of 3753 photographs (1720 shallow AC and 2033 deep AC images) were assigned to the training dataset, and 1302 photographs (509 shallow AC and 793 deep AC images) were held out for two internal testing datasets. In detecting shallow ACD in the internal hold-out testing dataset, the DL model achieved an AUC of 0.86 (95% CI, 0.83-0.90) with 80% sensitivity and 79% specificity. In the same testing dataset, the DL model also achieved better performance than the two glaucoma specialists (accuracy of 80% vs. accuracy of 74 and 69%). CONCLUSIONS: We proposed a high-performing DL model to automatically detect shallow ACD from overview anterior segment photographs. Our DL model has potential applications in detecting and monitoring shallow ACD in the real world. TRIAL REGISTRATION: http://clinicaltrials.gov , NCT04340635 , retrospectively registered on 29 March 2020.

Detecting glaucoma based on spectral domain optical coherence tomography imaging of peripapillary retinal nerve fiber layer: a comparison study between hand-crafted features and deep learning model.

PURPOSE: To develop a deep learning (DL) model for automated detection of glaucoma and to compare diagnostic capability against hand-craft features (HCFs) based on spectral domain optical coherence tomography (SD-OCT) peripapillary retinal nerve fiber layer (pRNFL) images. METHODS: A DL model with pre-trained convolutional neural network (CNN) based was trained using a retrospective training set of 1501 pRNFL OCT images, which included 690 images from 153 glaucoma patients and 811 images from 394 normal subjects. The DL model was further tested in an independent test set of 50 images from 50 glaucoma patients and 52 images from 52 normal subjects. A customized software was used to extract and measure HCFs including pRNFL thickness in average and four different sectors. Area under the receiver operator characteristics (AROC) curves was calculated to compare the diagnostic capability between DL model and hand-crafted pRNFL parameters. RESULTS: In this study, the DL model achieved an AROC of 0.99 [CI: 0.97 to 1.00] which was significantly larger than the AROC values of all other HCFs (AROCs 0.661 with 95% CI 0.549 to 0.772 for temporal sector, AROCs 0.696 with 95% CI 0.549 to 0.799 for nasal sector, AROCs 0.913 with 95% CI 0.855 to 0.970 for superior sector, AROCs 0.938 with 95% CI 0.894 to 0.982 for inferior sector, and AROCs 0.895 with 95% CI 0.832 to 0.957 for average). CONCLUSION: Our study demonstrated that DL models based on pre-trained CNN are capable of identifying glaucoma with high sensitivity and specificity based on SD-OCT pRNFL images.

Evaluation of retinal vasculature before and after treatment of children with obstructive sleep apnea-hypopnea syndrome by optical coherence tomography angiography.

PURPOSE: To evaluate the effect of adenoidectomy on the brains of children with obstructive sleep apnea-hypopnea syndrome (OSAS) through observation of the alteration of retinal perfusion by optical coherence tomography angiography (OCTA). METHODS: Sixty-two children with OSAS (124 eyes; 5.94 ± 1.64 years old; 53.2% boys) were enrolled in this study. Their retinal vascular network density indices, including vascular diameter (VD), vascular area density (VAD), vessel skeleton density (VSD), vessel perimeter index (VPI) on the macular superficial/deep capillary plexus (SCP/DCP), and the size of the foveal avascular zone (FAZ) were analyzed by OCTA before treatment and at 1 month after the adenoidectomy. RESULTS: After the adenoidectomy, the values of VD, VAD, VSD, and VPI in the SCP/DCP of the parafovea were significantly increased (p < 0.01). The FAZ was significantly diminished in the SCP (t = 4.50, p < 0.05) and increased in the DCP (t = - 4.43, p < 0.05). The peripapillary indices in the SCP/DCP were not significantly changed (p > 0.05). CONCLUSION: By improvement of hypoxia, the response of the parafoveal vessels was more sensitive than that of the peripapillary region in children with OSAS. Therefore, OCTA may be an ideal method to evaluate the changes of the retinal vascular system, which could be an effective parameter for the early evaluation of adenoidectomy on cases of pediatric OSAS.

Mutations of RagA GTPase in mTORC1 Pathway Are Associated with Autosomal Dominant Cataracts.

Cataracts are a significant public health problem with no proven methods for prevention. Discovery of novel disease mechanisms to delineate new therapeutic targets is of importance in cataract prevention and therapy. Herein, we report that mutations in the RagA GTPase (RRAGA), a key regulator of the mechanistic rapamycin complex 1 (mTORC1), are associated with autosomal dominant cataracts. We performed whole exome sequencing in a family with autosomal dominant juvenile-onset cataracts, and identified a novel p.Leu60Arg mutation in RRAGA that co-segregated with the disease, after filtering against the dbSNP database, and at least 123,000 control chromosomes from public and in-house exome databases. In a follow-up direct screening of RRAGA in another 22 families and 142 unrelated patients with congenital or juvenile-onset cataracts, RRAGA was found to be mutated in two unrelated patients (p.Leu60Arg and c.-16G>A respectively). Functional studies in human lens epithelial cells revealed that the RRAGA mutations exerted deleterious effects on mTORC1 signaling, including increased relocation of RRAGA to the lysosomes, up-regulated mTORC1 phosphorylation, down-regulated autophagy, altered cell growth or compromised promoter activity. These data indicate that the RRAGA mutations, associated with autosomal dominant cataracts, play a role in the disease by acting through disruption of mTORC1 signaling.

Quantitative assessment of changes in anterior segment morphology after argon laser peripheral iridoplasty: findings from the EARL study group.

IMPORTANCE: Argon laser peripheral iridoplasty (ALPI) could be effective in widening residual angle closure following laser peripheral iridotomy (LPI). BACKGROUND: We investigated changes in angle parameters following ALPI and its safety profile in this study. DESIGN: Retrospective, observational case series. PARTICIPANTS: The records from a single centre, of 36 patients (60 eyes) who underwent ALPI, for residual angle closure following LPI, were reviewed. METHODS: We analysed anterior chamber parameters in anterior segment optical coherence tomography (ASOCT) images using customized software pre- and post-ALPI. Paired t-test was used to compare changes. MAIN OUTCOME MEASURES: ASOCT parameters analysed included angle opening distance (AOD 500 and 750), trabecular iris surface area (TISA 500 and 750), anterior chamber width (ACW), anterior chamber volume (ACV), angle recess area (ARA), anterior chamber area (ACA), anterior chamber depth (ACD) and lens vault (LV). RESULTS: There was a mean increase in AOD 500 (0.05 vs. 0.16 mm, P < 0.001), AOD 750 (0.15 vs. 0.27 mm, P < 0.001), TISA 500 (0.010 vs. 0.038 mm2 , P < 0.001), TISA 750 (0.039 vs. 0.102 mm2 , P < 0.001), ACV (89.76 vs. 102.25 mm3 , P = 0.01), ARA 500 (0.015 vs. 0.033 mm2 , P < 0.001) and ARA 750 (0.044 vs. 0.088 mm2 , P < 0.001). There was no significant change in ACW, ACD, ACA and LV. Mean intraocular pressure (IOP) decreased post-ALPI (17.2 vs. 15.7 mmHg, P = 0.002). The mean follow-up duration was 2.1 years (range 0.5-5 years). CONCLUSIONS AND RELEVANCE: ALPI results in changes to the angle morphology and lowered IOP in eyes with residual angle closure. Our findings suggest a possible role for ALPI in eyes with residual angle closure following peripheral iridotomy.

Joint Detection and DOA Tracking with a Bernoulli Filter Based on Information Theoretic Criteria.

In this paper, we study the problem of the joint detection and direction-of-arrival (DOA) tracking of a single moving source which can randomly appear or disappear from the surveillance volume. Firstly, the Bernoulli random finite set (RFS) is employed to characterize the randomness of the state process, i.e., the dynamics of the source motion and the source appearance. To increase the performance of the detection and DOA tracking in low signal-to-noise ratio (SNR) scenarios, the measurements are obtained directly from an array of sensors and allow multiple snapshots. A track-before-detect (TBD) Bernoulli filter is proposed for tracking a randomly on/off switching single dynamic system. Secondly, since the variances of the stochastic signal and measurement noise are unknown in practical applications, these nuisance parameters are marginalized by defining an uninformative prior, and the likelihood function is compensated by using the information theoretic criteria. The simulation results demonstrate the performance of the filter.

Depth Discrimination for Low-Frequency Sources Using a Horizontal Line Array of Acoustic Vector Sensors Based on Mode Extraction.

Depth discrimination is a key procedure in acoustic detection or target classification for low-frequency underwater sources. Conventional depth-discrimination methods use a vertical line array, which has disadvantage of poor mobility due to the size of the sensor array. In this paper, we propose a depth-discrimination method for low-frequency sources using a horizontal line array (HLA) of acoustic vector sensors based on mode extraction. First, we establish linear equations related to the modal amplitudes based on modal beamforming in the vector mode space. Second, we solve the linear equations by introducing the total least square algorithm and estimate modal amplitudes. Third, we select the power percentage of the low-order modes as the decision metric and construct testing hypotheses based on the modal amplitude estimation. Compared with a scalar sensor, a vector sensor improves the depth discrimination, because the mode weights are more appropriate for doing so. The presented linear equations and the solution algorithm allow the method to maintain good performance even using a relatively short HLA. The constructed testing hypotheses are highly robust against mismatched environments. Note that the method is not appropriate for the winter typical sound speed waveguide, because the characteristics of the modes differ from those in downward-refracting sound speed waveguide. Robustness analysis and simulation results validate the effectiveness of the proposed method.

Forecastable and Guidable Bubble-Propelled Microplate Motors for Cell Transport.

Cell transport is important to renew body functions and organs with stem cells, or to attack cancer cells with immune cells. The main hindrances of this method are the lack of understanding of cell motion as well as proper transport systems. In this publication, bubble-propelled polyelectrolyte microplates are used for controlled transport and guidance of HeLa cells. Cells survive attachment on the microplates and up to 22 min in 5% hydrogen peroxide solution. They can be guided by a magnetic field whereby increased friction of cells attached to microplates decreases the speed by 90% compared to pristine microplates. The motion direction of the cell-motor system is easier to predict due to the cell being opposite to the bubbles.

Determinants of pupil diameters and pupil dynamics in an adult Chinese population.

BACKGROUND: To investigate the determinants of pupil diameter (PD), amplitude of pupil diameter change (PD-change) and speed of pupil constriction (SPC) using video anterior segment optical coherence tomography (AS-OCT) in a population-based sample of Chinese adults. METHODS: Chinese adults aged 40 to 80 years who were free from glaucoma were consecutively recruited from the population-based Singapore Chinese Eye Study. The SPC was measured by AS-OCT videography. Univariate and multivariate analyses were performed to examine the effects of demographic and ocular biometric factors (e.g., axial length [AL], anterior chamber depth [ACD], baseline PD, iris thickness at the area of the dilator muscle [ITDMR], iris area [IA], and iris bowing [IB]) on SPC, PD, and PD-change. RESULTS: A total of 266/302 (89.5 %) AS-OCT videos of eligible eyes were available for analysis. Among these subjects, 64.3 % were women, and the mean age (± standard deviation [SD]) was 56 ± 8.3 years. SPC was not associated with sex. In multiple regression analyses, SPC was independently associated with baseline PD (ß = 0.116, p = 0.006). Baseline PD was independently associated with ACD (ß = 0.341, p < 0.001), TISA 500 (ß = -4.513, p < 0.001), IA (ß = -2.796, p < 0.001), and ITDMR (ß = 6.573, p < 0.001). PD-change was independently associated with ACD (ß = 0.256, p < 0.001), IA (ß = -1.507, p < 0.001), IB (ß = 0.630, p = 0.011), and ITDMR (ß = 3.124, p < 0.001). CONCLUSIONS: Among normal eyes in an adult Chinese population, SPC was associated with larger baseline PD. Larger baseline PD and greater PD change form dark to light were associated with greater ACD, with smaller IA and thicker ITDMR.

Pretreatment anterior segment imaging during acute primary angle closure: insights into angle closure mechanisms in the acute phase.

OBJECTIVE: To describe anterior segment optical coherence tomography (ASOCT) parameters during acute primary angle closure (APAC) before therapeutic interventions and comparative analyses of biometric parameters of APAC eyes with fellow eyes. DESIGN: Prospective, comparative case series. PARTICIPANTS: Thirty-one consecutive patients with APAC. METHODS: All patients underwent ASOCT imaging of both eyes during the attack, before therapeutic interventions were administered. Custom software was used to measure anterior chamber depth (ACD), anterior chamber area (ACA), anterior chamber volume (ACV), iris curvature (I-Curv), iris area (I-Area), lens vault (LV), and angle opening distance (AOD750), trabecular iris space area (TISA750), and iris thickness (IT750) at 750 µm from the scleral spur. Multivariate logistic regression modeling using forward selection was used to determine the most important biometric variables associated with APAC compared with the fellow eye during the attack. MAIN OUTCOME MEASURES: Anterior segment biometric parameters associated with APAC. RESULTS: The mean age of the patients was 60.9±7.5 years, and 11 patients (35.5%) were male. The mean intraocular pressure was 3.8±9.2 mmHg in the APAC eye and 4.2±4.3 mmHg in the fellow eye before treatment (P <0.001). After adjustment for pupil diameter, APAC eyes had smaller ACD, ACA, ACV, I-Curv (all P <0.001), AOD750 (P = 0.037), TISA750 (P = 0.043), I-Area (P = 0.027), and IT750 (P = 0.002) and larger LV (P = 0.041) than fellow eyes. An optimal model consisting of 3 variables (pupil diameter, ACD, and I-Curv) explained 36.7% of the variance in APAC occurrence, with ACD accounting for 18.1% and I-Curv accounting for 14.1% of this variance. CONCLUSIONS: Shallower ACD and smaller I-Curv were the 2 main anterior segment biometric parameters associated with APAC during the attack. These findings present new insights into the anterior segment biometric parameters of APAC and fellow eyes before therapeutic interventions. Anatomic changes in the anterior segment explained only about one third of the variance in APAC occurrence, and the role of nonanatomic factors require further investigation.

[Water use characteristics and the mechanism of water uptake in two typical sand-fixing species in Mu Us sandy land]. / æ¯›ä¹Œç´ å ¸åž‹å›ºæ²™æ¤ç‰©çš„æ°´åˆ†åˆ©ç”¨ç‰¹æ€§ä¸Žå¸æ°´æœºåˆ¶.

Understanding water absorption mechanisms of sand-fixing plants is important for the rational establishment of plant community structures, thereby providing a scientific basis for desertification control and the efficient utilization of water resources in sandy areas. Based on the hydrogen and oxygen isotopic compositions of precipi-tation, soil water, xylem water, and groundwater, coupled with soil water-heat dynamics, annual water consumption characteristics of vegetation, using the multi-source linear mixing model (IsoSource), we analyzed the differences in water sources between Salix psammophila and Artemisia ordosica, during winter and the growing season. We further examined the effects of groundwater depth (2 m and 10 m), soil freezing-thawing, and drought on their water utilization to elucidate water absorption mechanisms of those species. The results showed that: 1) During soil freezing-thawing period (January to March), S. psammophila mainly utilized soil water in 60-120 cm depths below the frozen layer (69.1%). In the green-up season (April and May), soil water from the 0-60 cm layers could satisfy the water demand of S. psammophila (30.9%-87.6%). During the dry period of the growing season (June), it predominantly utilized soil water at the depth of 120-160 cm (27.4%-40.8%). Over the rainy season (July and September), soil water in 0-60 cm depths provided 59.8%-67.9% of the total water required. A. ordosica, with shallow roots, could not utilize soil water after complete freezing of root zone but could overwinter by storing water in rhizomes during autumn. During the growing season, it primarily relied on 0-40 cm soil layer (23.4%-86.8%). During the dry period, it mainly utilized soil water from 40-80 cm and 80-160 cm soil layers, with utilization rates of 14.6%-74.4% and 21.8%-78.2%, respectively. 2) With decreasing groundwater depth, vegetation shifted its water absorption depth upward, with water source of S. psammophila transitioning from 120-160 cm to 60-160 cm layers, while A. ordosica shifted water absorption depth from 80-160 cm to 0-40 cm. S. psammophila's utilization of soil water is influenced by transpiration, adopting an "on-demand" approach to achieve a balance between water supply and energy conservation, whereas A. ordosica tends to utilize shallow soil water, exhibiting a higher depen-dence on water sources from a single soil layer.

Development and evaluation of a large language model of ophthalmology in Chinese.

BACKGROUND: Large language models (LLMs), such as ChatGPT, have considerable implications for various medical applications. However, ChatGPT's training primarily draws from English-centric internet data and is not tailored explicitly to the medical domain. Thus, an ophthalmic LLM in Chinese is clinically essential for both healthcare providers and patients in mainland China. METHODS: We developed an LLM of ophthalmology (MOPH) using Chinese corpora and evaluated its performance in three clinical scenarios: ophthalmic board exams in Chinese, answering evidence-based medicine-oriented ophthalmic questions and diagnostic accuracy for clinical vignettes. Additionally, we compared MOPH's performance to that of human doctors. RESULTS: In the ophthalmic exam, MOPH's average score closely aligned with the mean score of trainees (64.7 (range 62-68) vs 66.2 (range 50-92), p=0.817), but achieving a score above 60 in all seven mock exams. In answering ophthalmic questions, MOPH demonstrated an adherence of 83.3% (25/30) of responses following Chinese guidelines (Likert scale 4-5). Only 6.7% (2/30, Likert scale 1-2) and 10% (3/30, Likert scale 3) of responses were rated as 'poor or very poor' or 'potentially misinterpretable inaccuracies' by reviewers. In diagnostic accuracy, although the rate of correct diagnosis by ophthalmologists was superior to that by MOPH (96.1% vs 81.1%, p>0.05), the difference was not statistically significant. CONCLUSION: This study demonstrated the promising performance of MOPH, a Chinese-specific ophthalmic LLM, in diverse clinical scenarios. MOPH has potential real-world applications in Chinese-language ophthalmology settings.

Swept source optical coherence tomography measurement of the iris-trabecular contact (ITC) index: a new parameter for angle closure.

PURPOSE: To evaluate the inter- and intra-observer agreement of measurement of the iris-trabecular contact (ITC) index, a measure of the degree of angle closure, using swept source optical coherence tomography (SSOCT, CASIA SS-1000, Tomey Corporation, Nagoya, Japan). METHODS: One randomly selected eye of 60 subjects was imaged under dark room conditions. The SSOCT 3-dimensional angle scan simultaneously obtains 128 radial scans of the anterior chamber for the entire circumference of the angle. Post-imaging analysis estimated the ITC index using in-built software. For intra-observer agreement for image grading, one examiner performed the grading twice in a masked fashion and random order after a 1-week interval. A second examiner graded images to assess inter-observer agreement for image grading. For intra-observer agreement for image acquisition, a single operator imaged patients twice. For inter-observer agreement for image acquisition, a single observer graded two sets of images acquired by two different operators on the same patient. Bland-Altman plots and 95 % limits of agreement (LOA) were reported. RESULTS: Study subjects were predominantly Chinese (54/60, 90 %) and female (42/60, 70 %), with a mean age of 65.5 years. The median ITC index for eyes with open angles (31/60) and closed angles was 20 % (95 % confidence interval [CI] - 13.6, 27.8) and 49 % (95%CI - 35.5, 69.2) respectively. The mean difference (95 % LOA) for intra-observer agreement for image grading and image acquisition were -0.8 % (-8.2, 6.5) and 0.6 % (-10.9, 9.7); corresponding inter- observer agreement were 0.1 % (-10, 10.1) and -0.3 % (-11.1, 10.5) respectively. CONCLUSIONS: The inter- and intra-observer agreement of the ITC index, as a measure of extent of angle closure using SSOCT, was good.

Assessment of trabecular meshwork width using swept source optical coherence tomography.

PURPOSE: Measurements of the angle width by ultrasound biomicroscopy or anterior segment optical coherence tomography are usually performed 500 µm from the scleral spur, as the anterior part of trabecular meshwork (TM) is assumed to lie within this distance. The aim of this study was to measure TM width using swept source optical coherence tomography (SS-OCT, CASIA SS-1000, Tomey Corporation, Nagoya, Japan), and to investigate factors influencing this measurement. METHODS: Participants underwent gonioscopy and SS-OCT imaging in the dark. High-definition SS-OCT images were corrected for refractive distortion; and customized software (ImageJ; National Institutes of Health, Bethesda, MD, USA) was utilized to measure TM width (distance between the scleral spur and Schwalbe's line). Linear regression analysis was performed to assess the relationship between TM width with demographic and angle parameters. RESULTS: One hundred and forty eight Chinese subjects were analyzed. The majority was female (62.4 %); the mean age was 59.2 ± 8.68 years. Identification of the scleral spur and Schwalbe's line with SS-OCT was possible in 590 (99.7 %) and 585 angle quadrants (98.8 %) respectively. TM width was wider in the inferior and superior quadrants (mean 889 [SD 138] and 793 [136] µm), compared to the nasal and temporal quadrants (712 [137] and 724 [115] µm, P<0.001). There was a difference in average TM width between open (789 [100]) and closed angle eyes (753 [86]) (P=0.048). There was no significant association between TM width and angle parameters, laterality, or demographic factors. CONCLUSIONS: In SS-OCT HD images, the mean TM width varied from 710 to 890 µm in the different quadrants of the eye, and the inferior quadrant TM was the widest compared to other quadrants.

Annotation-efficient learning for OCT segmentation.

Deep learning has been successfully applied to OCT segmentation. However, for data from different manufacturers and imaging protocols, and for different regions of interest (ROIs), it requires laborious and time-consuming data annotation and training, which is undesirable in many scenarios, such as surgical navigation and multi-center clinical trials. Here we propose an annotation-efficient learning method for OCT segmentation that could significantly reduce annotation costs. Leveraging self-supervised generative learning, we train a Transformer-based model to learn the OCT imagery. Then we connect the trained Transformer-based encoder to a CNN-based decoder, to learn the dense pixel-wise prediction in OCT segmentation. These training phases use open-access data and thus incur no annotation costs, and the pre-trained model can be adapted to different data and ROIs without re-training. Based on the greedy approximation for the k-center problem, we also introduce an algorithm for the selective annotation of the target data. We verified our method on publicly-available and private OCT datasets. Compared to the widely-used U-Net model with 100% training data, our method only requires ∼10% of the data for achieving the same segmentation accuracy, and it speeds the training up to ∼3.5 times. Furthermore, our proposed method outperforms other potential strategies that could improve annotation efficiency. We think this emphasis on learning efficiency may help improve the intelligence and application penetration of OCT-based technologies.

Self-propelled bioglass janus nanomotors for dentin hypersensitivity treatment.

Dentin hypersensitivity treatment is not always successful owing to the exfoliation of the blocking layer. Therefore, efficiently delivering a desensitization agent into the dental tubule is critical. Nanomotors are widely used as in vivo drug delivery systems owing to their strong power and good biocompatibility. Herein, we report a kind of self-propelled bioglass Janus nanomotor with a Pt motion unit (nBGs@Pt) for application in dentin hypersensitivity that was prepared via a simple sol-gel method and magnetron sputtering method, with an average size of 290 nm. The Pt layer as the power unit provided the dynamics to deliver the bioglass (desensitization agent). Using hydrogen peroxide as a fuel, the nBGs@Pt could automatically move in different media. In addition, the nBGs@Pt with a mesoporous structure demonstrated good hydroxyapatite formation performance. An in vitro dentin pressure model was used to verify the blocking ability of the nBGs@Pt in dentin tubules. The dynamics of the nBGs@Pt was sufficient to resist the outflow of dentin fluid and movement into the dentin tubules, with a blocking rate of 58.05%. After remineralization, the blocking rate could reach 96.07% and the formation of hydroxyapatite of up to 10 µm or more occurred. It is expected that this study will provide a simple and feasible new strategy for the painless treatment of dentin sensitivity.