Does language help make sense of the visual world? How important is it to actually see the world rather than having it described with words? These basic questions about the nature of intelligence have been difficult to answer because we only had one example of an intelligent system -- humans -- and limited access to cases that isolated language or vision. However, the development of sophisticated Vision-Language Models (VLMs) by artificial intelligence researchers offers us new opportunities to explore the contributions that language and vision make to learning about the world. We ablate components from the cognitive architecture of these models to identify their contributions to learning new tasks from limited data. We find that a language model leveraging all components recovers a majority of a VLM's performance, despite its lack of visual input, and that language seems to allow this by providing access to prior knowledge and reasoning.

PURPOSE: To assess the accuracy of intraoperative wavefront aberrometry (IWA) versus modern intraocular lens formulas in post-myopic laser vision correction (LVC) patients undergoing cataract surgery with capsular tension ring placement. METHODS: This is a retrospective chart review conducted at an academic outpatient center. All post-myopic LVC eyes undergoing cataract surgery with IWA from a single surgeon from 05/2017 to 12/2019 were included. All patients received a capsular tension ring (CTR). Mean numerical error (MNE), median numerical error (MedNE), and percentages of prediction error within 0.50D, 0.75D, and 1.00D were calculated for the above formulas. RESULTS: Twenty-seven post-myopic LVC eyes from 18 patients were included. In post-myopic LVC, MNE with Optiwave Refractive Analysis (ORA), Barrett True K (BTK), Haigis, Haigis-L, Shammas, SRK/T, Hill-RBF v3.0, and W-K AL-adjusted Holladay 1 were + 0.224, - 0.094, + 0.193, - 0.231, - 0.372, + 1.013, + 0.860, and + 0.630 (F = 8.49, p < 0.001). MedNE were + 0.125, - 0.145, + 0.175, + 0.333, + 0.333, + 1.100, + 0.880, and + 0.765 (F = 7.89, p < 0.001), respectively. BTK provided improved accuracy in both MNE (p < 0.001) and MedNE (p = .033) when compared to ORA in pairwise analysis. If the ORA vs. BTK-suggested IOL power were routinely selected, 30% and 15% of eyes would have projected hyperopic outcomes, respectively (p = 0.09). CONCLUSIONS: Our study suggests that in post-myopic LVC eyes undergoing cataract surgery with CTRs, BTK performed more accurately than ORA with regard to accuracy and yielded a lower percentage of eyes with hyperopic outcomes. Haigis, Haigis-L, and Shammas yielded similar results to ORA with regard to accuracy and percentage of eyes with hyperopic outcomes. On average, Shammas and Haigis-L suggested IOLs that would yield outcomes more myopic than expected when compared to BTK.

Previous studies have established migration patterns between the geographic location of physicians’ residency programs and their first post-training job. Our study explores the patterns of migration of pre-residency education locations with residency and post-residency geography. We analyzed responses to an annual survey administered between 2008 and 2013 to participants of the board examination review courses organized by Galderma Laboratories. Geography of high school (HS), medical school (MS), residency, and first job location were highly correlated. The Midwest and South retained the most residents from HS (70% each), whereas the West retained the lowest percent of residents from HS (33%). The West and Northeast exported about half of their HS graduates to the South and MW for residency. The South retained the largest proportion of its trainees post-residency (75%). Our data revealed that both HS and MS locations are closely related to residents’ ultimate employment locations. This information may be useful to training program directors and chairpersons as they manage recruitment and retention of trainees and faculty and may inform plans to address geographic workforce imbalances.

Purpose

Despite increasing electronic health record (EHR) adoption, perceptions of EHRs are negative among ophthalmologists due to concerns about productivity, costs, and documentation. The authors evaluated the effects of EHR adoption in an oculoplastics practice, which had not been previously studied.

Methods

Clinical volume, documentation time, time spent with patients, reimbursement, relative value units, and patient satisfaction were examined for 2 academic oculoplastics attendings between April 2018 and April 2019, with EHR implementation in September 2018.

Results

The mean number of patients seen in a half-day clinic was 31.8 versus 27.7 (p = 0.018) pre- and post-EHR implementation, respectively. EHR implementation had no effect on total monthly reimbursement (p = 0.88) or total monthly relative value units (p = 0.54). Average reimbursement (p = 0.004) and relative value units (p = 0.001) per patient encounter were significantly greater with EHR use. Patient satisfaction scores improved (p = 0.018). Mean physician time per patient increased from 6.4 to 9.0 minutes (p < 0.001). Mean documentation time per patient increased from 1.7 to 3.6 minutes (p < 0.001). Average patient wait times decreased by 9 minutes (p = 0.03) with EHR use. No scribes were used.

Conclusions

EHR implementation was associated with decreased patient volume without significant differences in total reimbursement. Although EHR adoption was associated with increased physician time devoted to patients and greater time expenditure on documentation, patients experienced decreased wait times. This suggests that EHR use streamlined the overall clinic flow without sacrificing physicians' time with the patient. The author's findings suggest that EHR implementation can be accomplished in an academic oculoplastics setting without negative impact on patient experience or reimbursement considerations.

Cellular differentiation requires both activation of target cell transcriptional programs and repression of non-target cell programs. The Myt1 family of zinc finger transcription factors contributes to fibroblast to neuron reprogramming in vitro. Here, we show that ztf-11 (Zinc-finger Transcription Factor-11), the sole Caenorhabditis elegans Myt1 homolog, is required for neurogenesis in multiple neuronal lineages from previously differentiated epithelial cells, including a neuron generated by a developmental epithelial-to-neuronal transdifferentiation event. ztf-11 is exclusively expressed in all neuronal precursors with remarkable specificity at single-cell resolution. Loss of ztf-11 leads to upregulation of non-neuronal genes and reduced neurogenesis. Ectopic expression of ztf-11 in epidermal lineages is sufficient to produce additional neurons. ZTF-11 functions together with the MuvB corepressor complex to suppress the activation of non-neuronal genes in neurons. These results dovetail with the ability of Myt1l (Myt1-like) to drive neuronal transdifferentiation in vitro in vertebrate systems. Together, we identified an evolutionarily conserved mechanism to specify neuronal cell fate by repressing non-neuronal genes.

PURPOSE: The purpose of this study was to compare the histopathologic inflammation and fibrosis of orbital adipose tissue in orbital inflammatory disease (OID) specimens. METHODS: In this retrospective cohort study, inflammation, and fibrosis in orbital adipose tissue from patients with thyroid-associated orbitopathy (TAO), granulomatosis with polyangiitis (GPA), sarcoidosis, nonspecific orbital inflammation (NSOI), and healthy controls were scored by 2 masked ocular pathologists. Both categories were scored on a scale of 0 to 3 with scoring criteria based on the percentage of specimens containing inflammation or fibrosis, respectively. Tissue specimens were collected from oculoplastic surgeons at 8 international centers representing 4 countries. Seventy-four specimens were included: 25 with TAO, 6 with orbital GPA, 7 with orbital sarcoidosis, 24 with NSOI, and 12 healthy controls. RESULTS: The mean inflammation and fibrosis scores for healthy controls were 0.0 and 1.1, respectively. Orbital inflammatory disease groups inflammation (I) and fibrosis (F) scores, formatted [I, F] with respective p -values when compared to controls, were: TAO [0.2, 1.4] ( p = 1, 1), GPA [1.9, 2.6] ( p = 0.003, 0.009), sarcoidosis [2.4, 1.9] ( p = 0.001, 0.023), and NSOI [1.3, 1.8] ( p ≤ 0.001, 0.018). Sarcoidosis had the highest mean inflammation score. The pairwise analysis demonstrated that sarcoidosis had a significantly higher mean inflammation score than NSOI ( p = 0.036) and TAO ( p < 0.0001), but no difference when compared to GPA. GPA had the highest mean fibrosis score, with pairwise analysis demonstrating a significantly higher mean fibrosis score than TAO ( p = 0.048). CONCLUSIONS: Mean inflammation and fibrosis scores in TAO orbital adipose tissue samples did not differ from healthy controls. In contrast, the more intense inflammatory diseases such as GPA, sarcoidosis, and NSOI did demonstrate higher histopathologic inflammation and fibrosis. This has implications in prognosis, therapeutic selection, and response monitoring in orbital inflammatory disease.

Purpose

To identify risk factors for the development of new-onset, postoperative diplopia following orbital decompression surgery based on patient demographics, clinical exam characteristics, radiographic parameters, and surgical techniques.

Methods

We conducted a multi-center retrospective chart review of patients who underwent orbital decompression for thyroid eye disease (TED). Patient demographics, including age, gender, smoking history, preoperative exophthalmometry, clinical activity score (CAS), use of peribulbar and/or systemic steroids, and type of orbital decompression were reviewed. Postoperative diplopia was determined at a minimum of 3 months postoperatively and before any further surgeries. Cross-sectional area ratios of each extraocular muscle to orbit and total fat to orbit were calculated from coronal imaging in a standard fashion. All measurements were carried out using PACS imaging software. Multivariable logistic regression modeling was performed using Stata 14.2 (StataCorp, College Station, TX).

Results

A total of 331 patients without preoperative diplopia were identified. At 3 months postoperatively, 249 patients had no diplopia whereas 82 patients developed diplopia. The average postoperative follow-up was 22 months (range 3-156) months. Significant preoperative clinical risk factors for postoperative diplopia included older age at surgery, proptosis, use of peribulbar or systemic steroids, elevated clinical activity score, and presence of preoperative compressive optic neuropathy. Imaging findings of enlarged cross-sectional areas of each rectus muscle to the overall orbital area also conferred a significant risk of postoperative diplopia. Regarding surgical factors, postoperative diplopia was more common among those undergoing medial wall decompression, bilateral orbital surgery, and balanced decompression, whereas endoscopic medial wall decompression was found to be relatively protective.

Conclusions

This study identifies risk factors associated with the development of diplopia following orbital decompression using multivariable data. This study demonstrates that several characteristics including age, clinical activity score, the cross-sectional muscle to orbit ratios, in addition to the type of orbital decompression surgery, are predictive factors for the development of new-onset postoperative diplopia.

The human brain has enormously complex cellular diversity and connectivities fundamental to our neural functions, yet difficulties in interrogating individual neurons has impeded understanding of the underlying transcriptional landscape. We developed a scalable approach to sequence and quantify RNA molecules in isolated neuronal nuclei from a postmortem brain, generating 3227 sets of single-neuron data from six distinct regions of the cerebral cortex. Using an iterative clustering and classification approach, we identified 16 neuronal subtypes that were further annotated on the basis of known markers and cortical cytoarchitecture. These data demonstrate a robust and scalable method for identifying and categorizing single nuclear transcriptomes, revealing shared genes sufficient to distinguish previously unknown and orthologous neuronal subtypes as well as regional identity and transcriptomic heterogeneity within the human brain.