Performance Analysis of Deep-Neural-Network-Based Automatic Diagnosis of Diabetic Retinopathy

Diabetic retinopathy (DR) is a human eye disease that affects people who are suffering from diabetes. It causes damage to their eyes, including vision loss. It is treatable; however, it takes a long time to diagnose and may require many eye exams. Early detection of DR may prevent or delay the vision loss. Therefore, a robust, automatic and computer-based diagnosis of DR is essential. Currently, deep neural networks are being utilized in numerous medical areas to diagnose various diseases. Consequently, deep transfer learning is utilized in this article. We employ five convolutional-neural-network-based designs (AlexNet, GoogleNet, Inception V4, Inception ResNet V2 and ResNeXt-50). A collection of DR pictures is created. Subsequently, the created collections are labeled with an appropriate treatment approach. This automates the diagnosis and assists patients through subsequent therapies. Furthermore, in order to identify the severity of DR retina pictures, we use our own dataset to train deep convolutional neural networks (CNNs). Experimental results reveal that the pre-trained model Se-ResNeXt-50 obtains the best classification accuracy of 97.53% for our dataset out of all pre-trained models. Moreover, we perform five different experiments on each CNN architecture. As a result, a minimum accuracy of 84.01% is achieved for a five-degree classification.

Prognostic Biomarkers in Uveal Melanoma: The Status Quo, Recent Advances and Future Directions

Uveal melanoma (UM) is the most common malignant intraocular tumour in the adult population. It is a rare cancer with an incidence of nearly five cases per million inhabitants per year, which develops from the uncontrolled proliferation of melanocytes in the choroid (≈90%), ciliary body (≈6%) or iris (≈4%). Patients initially present either with symptoms like blurred vision or photopsia, or without symptoms, with the tumour being detected in routine eye exams. Over the course of the disease, metastases, which are initially dormant, develop in nearly 50% of patients, preferentially in the liver. Despite decades of intensive research, the only approach proven to mildly control disease spread are early treatments directed to ablate liver metastases, such as surgical excision or chemoembolization. However, most patients have a limited life expectancy once metastases are detected, since there are limited therapeutic approaches for the metastatic disease, including immunotherapy, which unlike in cutaneous melanoma, has been mostly ineffective for UM patients. Therefore, in order to offer the best care possible to these patients, there is an urgent need to find robust models that can accurately predict the prognosis of UM, as well as therapeutic strategies that effectively block and/or limit the spread of the metastatic disease. Here, we initially summarized the current knowledge about UM by compiling the most relevant epidemiological, clinical, pathological and molecular data. Then, we revisited the most important prognostic factors currently used for the evaluation and follow-up of primary UM cases. Afterwards, we addressed emerging prognostic biomarkers in UM, by comprehensively reviewing gene signatures, immunohistochemistry-based markers and proteomic markers resulting from research studies conducted over the past three years. Finally, we discussed the current hurdles in the field and anticipated the future challenges and novel avenues of research in UM.

Refractive Outcomes of Patients Treated for Retinopathy of Prematurity

Background: Retinopathy of prematurity (ROP) is a disease where the blood vessels in the eye fail to develop appropriately in infants born prematurely. Two effective treatments for ROP include laser photocoagulation therapy and intravitreal bevacizumab (IVB). Laser therapy has been linked to causing high myopia in children post-treatment. We hypothesize that patients treated with IVB alone or in combination with laser will have less myopia development than patients treated with laser therapy alone. Methods: Patient demographics, treatment details, refractive data at 6-9 months and 3-4 years, the occurrence of strabismus at 3 years, and the most recent vision data were collected from 133 ROP patients. Patients not treated at IU health and those lost to follow-up, or deceased before both eye exams, were excluded from the study. Quantitative analysis was used to compare the refractive error, strabismus, and vision outcomes between the three treatment groups. A linear regression model was used to analyze the relationship between the number of laser spots applied and refractive error. Results: Refractive outcomes at 6-9 months and 3-4 years, occurrence of strabismus, and vision outcomes were statistically similar between the three treatment groups. However, the laser group had the most occurrences of high myopia. We also observed a 0.002 unit decrease in refractive error, reported at 6–9-months, with each laser spot applied (p<0.001). This may be due to the influence of outliers because no significant relationship was seen at the 3–4 year exam. Conclusion and Potential Impact: There was no difference in outcomes among patients treated with IVB, laser, or a combination of both, with the exception of more myopic outliers in the laser-only group. We can therefore assume that ROP patients who have received one of these three treatments had developed differences in myopia independent of treatment modality.

Hypertensive Retinopathy Is Associated With Worse Cognitive Function in Women: The SWAN Study

Based on the 2018 American Academy of Neurology guideline, the prevalence of mild cognitive impairment (CI) increases from 6.7% at 60-64 years to 25.2% at 80-84 years. There is interest to identify potential biomarkers in the retina for CI and dementia. The aims of this analysis was to test whether hypertensive retinopathy (HR) was associated with cognitive function using data from the Study of Women’s Health Across the Nation (SWAN), Michigan cohort. SWAN, launched in 1996/97, is a longitudinal study of women traversing midlife and into late adulthood. Starting in 2000, cognitive function tests were administered: East Boston Memory Test immediate (EBMTi) and 10-minute delay (EBMTd) for verbal episodic memory; digit span backwards (DSB) for working memory; and symbol digits modalities test (SDMT) for perception speed, motor speed, and visual scanning. Z-scores were calculated for EBMTi, EBMTd, DSB and SDMT and averaged at follow-up visit 15 (2015/16). Eye exams were performed on 255 women (66 + 2.6 years) at follow-up visit 16 (2016/17). HR was based on presence/absence of arteriovenous nicking. Logistic regression showed a statistically significant association of lower average cognitive Z-score with HR (p-value 0.03, beta=-0.21 [95% confidence interval: -0.40 to -0.02]) adjusting for measured hypertension or anti-hypertensive drugs, years of measured hypertension, race, education, and age. Preliminary results indicate that HR is associated with lower cognitive test scores in women in their 60s-70s. This association between a simple ophthalmic exam finding of systemic hypertension, i.e., arteriovenous nicking, and lower cognitive function is consistent with a cerebrovascular mechanism of CI.

Amblyopia and Routine Eye Exam in Children: Parent’s Perspective

Amblyopia is a reduced best-corrected visual acuity of one or both eyes that cannot be attributed to a structural abnormality; it is a functional reduction in the vision of an eye caused by disuse during a critical period of visual development. It is considered the leading cause of visual defects in children. With early diagnosis and treatment, children with amblyopia can significantly improve their vision. However, if it is neglected and not treated during childhood, unfortunately, it permanently decreases vision. Therefore, prevention, detection, and treatment largely depend on parents. This article explores parents’ perspectives on amblyopia and routine examination of their children’s eyes. A cross-sectional study used an electronic questionnaire consisting of five main sections to assess the level of awareness of amblyopia among parents. As a result, a total of 325 participants were included in our analysis. 209 (64.3%) were mothers, and 116 (35/7%) were fathers. The age groups were 35–50 years of age (61.5%), 20–34 years (23.4%), and older than 50 years (15%). Participants with a history of amblyopia numbered 23 (7.1%), and 39 had an amblyopic child (12%). A good awareness level of amblyopia among parents was found in only 10 (3%) participants, a fair awareness level in 202 (62%), and 113 (35%) participants were classified as having a poor awareness level of amblyopia. Only 13.8% of the parents took their children for yearly routine eye exams, while the majority (72%) took their children only if they had a complaint, and 14.2% took them for eye checkups only before school entry. In conclusion, parents’ awareness of amblyopia in Tabuk City, KSA, was low. In addition, a limited proportion of parents reported consistently taking their children for routine eye exams. Therefore, raising awareness should be considered in public education regarding the disease.

Knowledge about juvenile idiopathic arthritis-associated uveitis: more frequent reminders are associated with higher patient and family uveitis knowledge

Background Chronic anterior uveitis is a sight-threatening complication of juvenile idiopathic arthritis (JIA) and a primary contributor to long-term morbidity in people with JIA. Levels of knowledge about uveitis among JIA patients and their parents are unknown. A survey of JIA patients and parents was conducted to assess knowledge about uveitis complications and recommended screening. Methods A survey was developed consisting of six demographic questions, six arthritis/uveitis history questions, and nine uveitis knowledge questions. The survey was administered to JIA patients age 14 and older and parents of patients with JIA at three pediatric rheumatology practices and online through the Patients, Advocates, and Rheumatology Teams Network for Research and Service (PARTNERS) network. ANOVA, chi-square and Fisher’s exact tests were used to look for relationships between survey questions and demographic variables. Results Thirty-three patients and 111 parents completed the survey. Overall, 17.4% reported a history of uveitis, and 89.6% had heard of uveitis. The mean composite knowledge score was 6.46 ± 2.6 out of 9. Patients and parents with a history of uveitis had higher composite knowledge scores than their counterparts without a uveitis history (p = 0.01 and p < 0.01, respectively). Parents whose rheumatologist reminded them about eye exams at every visit had higher knowledge of the risk of blindness (p = 0.04), the risk for uveitis when arthritis is controlled (p = 0.02), the need for ongoing eye exams when off of medications (p = 0.01), and had a higher overall score (p = 0.02) than those who were reminded at some visits or not at all. Conclusions JIA patients and parents report variable levels of knowledge regarding uveitis complications and recommended screening. Frequent discussion between the rheumatology provider and family about uveitis screening is associated with higher uveitis knowledge. Incorporating detailed and frequent education about uveitis into rheumatology clinic appointments may improve early uveitis detection and visual outcomes.

Using Machine Learning to Capture Quality Metrics from Natural Language: A Case Study of Diabetic Eye Exams

Background and Objective The prevalence of value-based payment models has led to an increased use of the electronic health record to capture quality measures, necessitating additional documentation requirements for providers. Methods This case study uses text mining and natural language processing techniques to identify the timely completion of diabetic eye exams (DEEs) from 26,203 unique clinician notes for reporting as an electronic clinical quality measure (eCQM). Logistic regression and support vector machine (SVM) using unbalanced and balanced datasets, using the synthetic minority over-sampling technique (SMOTE) algorithm, were evaluated on precision, recall, sensitivity, and f1-score for classifying records positive for DEE. We then integrate a high precision DEE model to evaluate free-text clinical narratives from our clinical EHR system. Results Logistic regression and SVM models had comparable f1-score and specificity metrics with models trained and validated with no oversampling favoring precision over recall. SVM with and without oversampling resulted in the best precision, 0.96, and recall, 0.85, respectively. These two SVM models were applied to the unannotated 31,585 text segments representing 24,823 unique records and 13,714 unique patients. The number of records classified as positive for DEE using the SVM models ranged from 667 to 8,935 (2.7–36% out of 24,823, respectively). Unique patients classified as positive for DEE ranged from 3.5 to 41.8% highlighting the potential utility of these models. Discussion We believe the impact of oversampling on SVM model performance to be caused by the potential of overfitting of the SVM SMOTE model on the synthesized data and the data synthesis process. However, the specificities of SVM with and without SMOTE were comparable, suggesting both models were confident in their negative predictions. By prioritizing to implement the SVM model with higher precision over sensitivity or recall in the categorization of DEEs, we can provide a highly reliable pool of results that can be documented through automation, reducing the burden of secondary review. Although the focus of this work was on completed DEEs, this method could be applied to completing other necessary documentation by extracting information from natural language in clinician notes. Conclusion By enabling the capture of data for eCQMs from documentation generated by usual clinical practice, this work represents a case study in how such techniques can be leveraged to drive quality without increasing clinician work.

Major cardiorespiratory events do not increase after immunizations, eye exams, and other stressors in most very low birth weight infants

BACKGROUND: Increased cardiorespiratory events with bradycardia and oxygen desaturation have been reported in very low birthweight (VLBW) infants following stressors such as immunizations. These events are difficult to quantify and may be mild. Our group developed an automated algorithm to analyze bedside monitor data from NICU patients for events with bradycardia and prolonged oxygen desaturation (BDs) and used this to compare BDs 24 hours before and after potentially stressful interventions. METHODS: We included VLBW infants from 2012–2017 with data available around at least one of four interventions: two-month immunizations, retinopathy of prematurity (ROP) examinations, ROP therapy, and inguinal hernia surgery. We used a validated algorithm to analyze electrocardiogram heart rate and pulse oximeter saturation data (HR, SpO2) to quantify BD events of HR < 100 beats/minute for≥4 seconds with oxygen desaturation < 80%SpO2 for≥10 seconds. BDs were analyzed 24 hours before and after interventions using Wilcoxon rank-sum tests. RESULTS: In 354 of 493 (72%) interventions, BD frequency stayed the same or decreased in the 24 hours after the event. An increase of at least five BD’s occurred in 17/146 (12%) after immunizations, 85/290 (29%) after ROP examinations, 4/33 (12%) after ROP therapy, and 3/25 (12%) after hernia surgery. Infants with an increase in BDs after interventions had similar demographics compared to those without. More infants with an increase in BDs following immunizations were on CPAP or caffeine than those without. CONCLUSIONS: Most VLBW infants in our cohort had no increase in significant cardiorespiratory events in the 24 hours following potentially stressful interventions.