Publications

PURPOSE: Accurate assessment and surveillance of retinoblastoma (RB) require more efficient and objective measurements. This study aims to develop an artificial intelligence (AI) system, named RB-Care, for automatic classification and quantitative assessment of RB.

METHODS: A total of 3730 wide-field fundus images were included for the development and validation of 2 models in RB-Care. The first model was trained to automatically classify the images into "normal," "unseeded RB," and "seeded RB." The second model performed quantitative assessment on unseeded RB by detecting and segmenting tumors and optic discs.

RESULTS: The classification model of RB-Care can accurately classify fundus images into 3 categories with an accuracy of 0.9734 and an area under the curve (AUC) of 0.9970. The segmentation model can make precise boundary detection and quantitative measurement on tumors and optic discs, achieving mean Intersection over Union (mIoU) of 0.9670 and Dice similarity coefficient (DSC) of 0.9780 for tumor segmentation, and mIoU of 0.9999 and DSC of 0.9999 for optic disc segmentation, which reaches a comparable level with ophthalmologists.

CONCLUSIONS: The RB-Care achieved excellent performance in both RB classification and segmentation. Consequently, the tumor size and the distance between tumor and optic disc can be quantified, which provides an objective measurement tool for quantitative assessment and surveillance of RB in clinical settings.

TRANSLATIONAL RELEVANCE: Developing a clinically relevant technologies for objective quantitative assessment of RB.

The visual system takes advantage of redundancy in the world by extracting summary statistics, a phenomenon known as ensemble perception. Ensemble representations are formed for low-level features like orientation and size and high-level features such as facial identity and expression. Whereas recent research has shown that the visual system forms intact ensemble representations even when faces are partially occluded via solid bars, how ensemble perception is impacted with the addition of naturalistic objects such as face masks or sunglasses is largely unknown. To investigate this, we conducted a series of experiments using continuous report tasks in which faces (either varying in identity or expression) were partially occluded with a surgical mask or sunglasses and participants had to report the average face using a face wheel. We found evidence that participants could still accurately extract the average even when a significant portion of it was occluded with either face masks or sunglasses. In a second experiment, however, we found performance was worse when the face wheel was variable trial to trial. Thus part of the preservation of performance in occlusion arises from the visual system learning the features of the particular face wheel being used. Overall, our results suggest that the visual system is able to establish robust ensemble representations for faces with naturalistic occlusions, but that robustness appears to be supported at least partially by learning information about the particular features that are informative for a given set of faces.

PURPOSE: Glaucoma requires regular visual field (VF) assessments. Eyecatcher 3.0 uses novel "smart glasses" hardware to provide a lightweight, low-cost solution, designed for use while unsupervised. This study aimed to determine the feasibility of using Eyecatcher for VF home-monitoring.

METHODS: Eyecatcher 3.0 consists of a smartphone, smart glasses, and wireless clicker. Functionally, it attempts to mimic the Humphrey Field Analyzer (HFA; - same task-instructions, stimuli, and outputs, but smaller field of view and luminance range). Five patients with glaucoma used Eyecatcher to test themselves at home for 3 months (both eyes, monocular, once-per-fortnight). Results from a reduced 24-2 grid were compared to HFA data collected in the clinic, and to normative Eyecatcher data collected from 76 normally sighted young adults. A subset of normally sighted participants (n = 16) also underwent two additional sessions of follow-up testing to assess repeatability. Usability was assessed via questionnaires.

RESULTS: All Eyecatcher tests were completed successfully (100%). There was reasonable agreement with the HFA in terms of mean deviation (MD; r = 0.85, P < 0.001) and observed pattern of loss. The HFA exhibited somewhat better repeatability than Eyecatcher (MD Coefficient of Repeatability = 2.9, 95% confidence interval [CI] = 2.1-4.1 decibels [dB] for HFA, vs. 3.9, 95% CI = 2.8-6.1 dB for Eyecatcher), although this difference was not statistically significant. Average Eyecatcher test duration was 6.5 minutes (both eyes). Patients generally rated the Eyecatcher as easy-to-use, although specific concerns were raised by some individuals.

CONCLUSIONS: Smart glasses may provide a feasible means of VFs home-monitoring. Eyecatcher yielded similar sensitivity values to the HFA, and most participants found the lightweight smart glasses acceptable to use. Further research is needed to establish diagnostic accuracy and clinical utility.

TRANSLATIONAL RELEVANCE: Validation of a new method of glaucoma home monitoring.

UNLABELLED: The mandible presents morphological variations, even in individuals without syndromes. This variability will determine different skeletal sagittal patterns, generally classified as Class I, II or III. The anatomical position of the mandibular canal has been investigated in different skeletal patterns, often using cone-beam computed tomography (CBCT) images, for diagnostic or surgical planning purposes.

AIM: The aim of this study is to perform a three-dimensional analysis of the position of the mandibular canal (MC) in adults with Class I, II and III skeletal patterns, by means of segmentation and 3D measurements on CBCT images.

MATERIALS AND METHOD: 75 CBCT images were obtained from a secondary database, and 3D analysis was performed using ITK-SNAP and 3D Slicer software. The 3D evaluation consisted of determining the orientation of the position of the mandible, segmentation of the mandible and the MC, creating 3D models, and establishing anatomical landmarks. Vertical (supero-inferior, SI), transverse (mediolateral, RL,) and 3D measurements were performed.

RESULTS: The position of the MC is modified according to the skeletal pattern and by morphological factors of the mandible such as sex and gonial angle. The proximity of the MC to the oblique line is smaller in the SI direction in Class III, and the position of the MC is associated with variation in the gonial angle. It may be closer to the cortical lingual in the central region.

CONCLUSION: The mandibular canal position should be considered in tomographic evaluation during diagnosis and therapeutic planning of mandible surgeries, especially in cases of sagittal ramus osteotomy.

We present data showing that the urinary metabolic ratio (MR) of metabolite to parent drug can be used to estimate the drug-drug interactions (DDIs) of pain management and substance abuse treatment medications with other coadministered drugs. We quantitatively measure 18 drugs and their phase I metabolites and monitor the effects of 14 interfering drugs on their MRs. The 18 drugs include dextromethorphan, oxycodone, hydrocodone, tramadol, morphine, buprenorphine, fentanyl, clonazepam, alprazolam, quetiapine, carisoprodol, tapentadol, ketamine, methadone, impramine, and amitriptyline. The 14 interfering drugs include fluoxetine, paroxetine, bupropion, citalopram, sertraline, venlafaxine, duloxetine, risperidone, trazodone, aripiprazole, cyclobenzaprine, amphetamine, and tetrahydrocannabinol. Some of these interfering drugs are inhibitors of either the CYP2D6, CYP3A4/5, or CYP2C19 pathways. By using the urinary MR of metabolite/parent drug, we observed patterns of inhibition and enhancement due to DDIs. Using the MR reference intervals of the 18 drug pairs established in an earlier study, and the current DDI system, we can alert providers of unusual metabolism caused by DDIs. This will help providers do better prescribing or review more closely all medications and supplements patients are taking, thus avoiding underdosing or potential medication adverse reactions.

We present data that show that quantitative urine drug concentrations obtained from individuals monitored for drug compliance as part of their participation in chronic opioid or substance abuse treatment can be used to quantify drug metabolism. We quantitatively monitor 18 drugs and their Phase 1 metabolite. These drugs were dextromethorphan, morphine, oxycodone, hydrocodone, quetiapine, tapentadol, tramadol, buprenorphine, clonazepam, fentanyl, imipramine, ketamine, carisoprodol, alprazolam, methadone, and amitriptyline. By using the ratio of metabolite/parent drug (prescribed medication), the expected or reference values for 18 drugs were obtained. Ratio values outside of this reference range could be considered to be caused by genetic metabolizing variants, drug-drug interactions, age, or deception. Alerting providers of the variance in metabolism from the expected norm might reduce overdosing or underdosing patients.

PURPOSE: To assess choroidal vasculature in bilateral age-related macular degeneration (AMD) patients, with neovascular AMD in one eye and dry AMD in the other, using an innovative three-dimensional (3D) algorithm.

METHODS: This retrospective cross-sectional study included 30 patients with asymmetric bilateral AMD. All patients underwent clinical exams and swept-source optical coherence tomography using PlexElite-9000 (Carl Zeiss Meditec, Dublin, CA) scans centered on the fovea. The 3D choroidal vasculature was segmented using our validated deep learning algorithm. These 3D maps were divided into five sectors, with the three largest vessels in each sector selected for mean choroidal vessel diameter and inter-vessel distance (IVD) measurements. Volumetric choroidal thickness (ChT) and choroidal vascularity index (CVI) were calculated. Analysis was performed using a linear mixed model.

RESULTS: This study analyzed 60 eyes from 30 patients, with a mean age of 78.63 ± 8.01 years. Neovascular AMD eyes had significantly higher mean LogMar visual acuity (0.63 ± 0.60 vs. 0.20 ± 0.18, P < 0.001). Neovascular eyes had higher average vessel diameter (253.711 ± 20.777 vs. 232.641 ± 28.249 micrometers, P < 0.001), significant in central, temporal, and superior sectors (P < 0.05). IVD was lower in neovascular AMD (209.563 ± 21.074 vs. 213.867 ± 30.684 µm, P = 0.515). Average ChT was lower in neovascular AMD (193.939 ± 49.627 vs. 198.245 ± 53.786 µm, P = 0.809), significant in the inferior sector (P = 0.011). Additionally, CVI was higher in neovascular AMD (0.365 ± 0.033 vs. 0.353 ± 0.035, P = 0.031), particularly in inferior, superior, and central sectors (P < 0.05).

CONCLUSIONS: Eyes with neovascular AMD demonstrated higher choroidal vessel diameter, ChT, and CVI and decreased IVD compared to dry AMD eyes, possibly related to vessel congestion and inflammation in eyes with active choroidal neovascularization.

Not-In-My-Backyard (NIMBY) facilities, often characterized by their negative externalities, are frequently opposed or resisted by the public. To examine the impact of information interventions on public acceptance of NIMBY facilities, a behavioral investigation experiment consisting of a survey of 100 college students before and after viewing negative videos of either environmentally polluting or psychologically excluding types of NIMBY facilities was conducted to assess the efficacy of these interventions in shaping public perceptions and attitudes. Differences in respondent's attitudes toward the types of NIMBY facilities demonstrated that the information intervention affected perceptions and attitudes toward NIMBY facilities. Changes in response variables for perceived risk, perceived benefits, self-efficacy, positive emotion label, social environment, and public acceptance showed that environmentally polluting NIMBY facilities elicited a more negative response than psychologically excluding NIMBY facilities. Post-intervention, significant differences emerged across all six dimensions, and attitudes toward environmentally polluting NIMBY facilities became more negative than those toward psychologically excluding NIMBY facilities, with a significant increase in the perceived risk and a significant decrease in the perceived benefit. The study demonstrated that information interventions influenced attitudes toward NIMBY facilities, and that this influence differed between the two facility types.

Continuous tracking is the recently developed psychophysical technique for efficiently estimating human visual temporal characteristics. The standard version of the task, referred to as position tracking (PT), asks participants to track the location of a continuously moving target by a motor response (e.g., mouse movement). Some studies have also used a variant method, attribute tracking (AT), which requires participants to track and reproduce a continuously changing attribute (e.g., luminance) of the target instead of position. For both PT and AT, the temporal dynamics of the entire system from vision to action can be estimated from the cross-correlogram (CCG) of the trajectory between the stimulus and response. The similarities and differences in CCG between PT and AT, however, remain elusive but were examined in this study. Experiment 1 compared the two CCGs using luminance-defined circular patches, color-contrast-defined patches, and luminance-defined patches with various spatial frequencies. The results indicate that the PT response was faster and less affected by the stimulus variables than the AT response. Experiment 2 showed that these differences could be reduced by making the visuomotor mapping of PT less direct by reversing the motor response direction and by making the local stimulus change magnitude comparable between PT and AT. The comparison with the traditional reaction time measures (Experiment 3) further showed that the peak latency of CCG from PT aligned better with the simple reaction time, whereas that from AT aligned better with the choice reaction time. These results indicate that CCG is more sluggish for AT than for PT because AT includes the process of identifying the stimulus content (attribute change direction) and mapping it to a motor response arbitrarily specified by the experimenter, and because the effective stimulus change magnitude for AT is often weaker than that for PT. These findings provide a clearer understanding of the meaning of CCGs measured by the two types of continuous tracking tasks.