Real-Time Risk Score for Glaucoma Mass Screening by Spectral Domain Optical Coherence Tomography: Development and Validation

Artificial intelligence in the diagnosis of glaucoma and neurodegenerative diseases

Artificial Intelligence is a rapidly expanding field within computer science that encompasses the emulation of human intelligence by machines. Machine learning and deep learning - two primary data-driven pattern analysis approaches under the umbrella of artificial intelligence - has created considerable interest in the last few decades. The evolution of technology has resulted in a substantial amount of artificial intelligence research on ophthalmic and neurodegenerative disease diagnosis using retinal images. Various artificial intelligence-based techniques have been used for diagnostic purposes, including traditional machine learning, deep learning, and their combinations. Presented here is a review of the literature covering the last 10 years on this topic, discussing the use of artificial intelligence in analysing data from different modalities and their combinations for the diagnosis of glaucoma and neurodegenerative diseases. The performance of published artificial intelligence methods varies due to several factors, yet the results suggest that such methods can potentially facilitate clinical diagnosis. Generally, the accuracy of artificial intelligence-assisted diagnosis ranges from 67-98%, and the area under the sensitivity-specificity curve (AUC) ranges from 0.71-0.98, which outperforms typical human performance of 71.5% accuracy and 0.86 area under the curve. This indicates that artificial intelligence-based tools can provide clinicians with useful information that would assist in providing improved diagnosis. The review suggests that there is room for improvement of existing artificial intelligence-based models using retinal imaging modalities before they are incorporated into clinical practice.

Comparative Evaluation of Fundus Image Interpretation Accuracy in Glaucoma Screening Among Different Physician Groups

Purpose

To examine the variability in glaucoma screening using fundus images among physicians, including non-ophthalmologists.

Patients and Methods

Sixty-nine eyes from 69 patients, including 25 eyes with glaucoma, were included from the Jikei University Hospital from July 2019 to December 2022. Fundus images were captured using TRC-NW8 (Topcon Corporation, Tokyo, Japan), and were interpreted by 10 non-ophthalmologists, 10 non-specialist ophthalmologists, and 9 specialists for diagnostic accuracy. We analyzed differences in diagnostic accuracy among the three groups. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and Kappa coefficient were compared, using the Kruskal-Wallis test followed by a post hoc Dunn's test.

Results

The sensitivity and specificity were 0.22 and 0.92 for non-ophthalmologists, 0.49 and 0.83 for non-specialist ophthalmologists, and 0.68 and 0.87 for specialists, respectively. Both specialists and non-specialist ophthalmologists showed significantly higher sensitivity than non-ophthalmologists (Dunn's test, P<0.001 and P=0.031). There was no significant difference in specificity among the three groups (Kruskal-Wallis test, P=0.086). The PPV did not differ significantly between the groups (Kruskal-Wallis test, P=0.108), while the NPV was significantly higher in specialists compared to non-ophthalmologists (Dunn's test, P<0.001). Specialists also had a significantly higher Kappa coefficient than non-ophthalmologists and non-specialist ophthalmologists (Dunn's test, P<0.001 and P=0.024).

Conclusion

Diagnostic accuracy varied significantly based on the physician's background.

Basic assessment on adding platelet measurement to legal health checkup in Japan: A cross-sectional and 20-year longitudinal study

Background

In Japan, health checkups for workers are legally compulsory. Considering legal health checkup items are important for Japanese workers' health problems. To date, the legal health checkup items for blood cell counts include only red blood cell counts and hemoglobin but not platelet counts. This study aimed to investigate the significance of measuring platelets in workers by showing the association between the FIB-4 index (FIB-4), which can be easily calculated from factors including platelet counts and viral hepatitis infection.

Method

Both cross-sectional and longitudinal analyses were conducted on the comprehensive medical examinations of male workers. In fiscal year (FY) 2019, a logistic regression model was applied to 12,918 examinees. For 13,459 examinees (mean age = 47.5 ± 9.3 SD), FY2000 was set to be followed until FY2019. A total of 149,956 records between FY2000 and FY2019 were analyzed cross-sectionally, and 8,038 men who were consecutively examined to FY2019 at the longest were analyzed longitudinally. Receiver operating characteristic (ROC) curve–area under the ROC curve (ROC–AUC) and Cox proportional methods were used to examine the association between platelet-related indices and viral hepatitis infection.

Results

Logistic regression showed that the risk of FIB-4 ≥ 2.67 was mostly associated with hepatitis C virus antibody (HCVAb) positivity [odds ratio (OR) = 2.51, 95% confidence interval (CI) = 1.08–5.86], while negatively associated with body mass index (BMI) (OR = 0.54, 95% CI = 0.30–0.97), and not associated with the presence of fatty liver. To detect HVC Ab positivity, ROC–AUC showed more effectiveness in FIB-4 than in the AST/ALT ratio (0.776, 95% CI = 0.747–0.773 vs. 0.552; 95% CI = 0.543–0.561). The Cox analysis showed that the risk of FIB-4 ≥ 2.67 was closely associated with hepatitis B virus surface antigen (HBsAg) [hazard ratio (HR) = 3.1, 95% CI = 2.0–4.6] and HCV Ab positivity (HR = 3.2, 95% CI = 2.0–5.0).

Conclusion

Our results suggest that it might be worth considering that usage of information on platelets in legal health checkups could be some help not to overlook workers with hepatitis virus carriers as a complementary countermeasure, although further investigations are needed into its practical application.

Significance of Fib4 index as an indicator of alcoholic hepatotoxicity in health examinations among Japanese male workers: a cross-sectional and retrospectively longitudinal study

BACKGROUND

Fib4 index (Fib4) is clinically used as a noninvasive marker of liver fibrosis. In this study, we aimed to preliminarily investigate whether Fib4 can be used to detect individuals who need assessment for alcoholic liver disease (ALD) in the general population by clarifying the detailed association of Fib4 with alcohol consumption and gamma-glutamyl transferase (GGT) among male workers.

METHODS

We analyzed data sets on the comprehensive medical examinations of male workers as cross-sectional and retrospectively longitudinal studies. We enrolled 10 782 males (mean age: 52.2 ± 10.2 years) in FY2019 and 7845 males (mean follow-up: 12.6 ± 6.7 years) who could be consecutively followed up for 20 years from FY2000 to FY2019. Data were evaluated using logistic regression and COX proportional analysis.

RESULTS

In the cross-sectional setting, the rate of Fib4 ≥ 2.67 in heavy drinkers (≥ 40 g of ethanol/day) was increased dose dependently in those over 65 years old, and that of body mass index ≥ 30 kg/m² was increased in those over 60 years old, but not in those with fatty liver. The odds ratio (OR) (95% confidence interval [CI]) for heavy drinking was 4.30 (95% CI = 1.90-9.72), and GGT ≥ 200 IU/L was considerably high (OR = 29.05 [95% CI = 17.03-49.56]). In the longitudinal setting, heavy drinkers and those with GGT ≥ 200 IU/L at 10 years after the baseline showed an increased risk for Fib4 ≥ 2.67 (hazard ratio = 2.17 [95% CI = 1.58-2.98] and 7.65 [95% CI 5.26-11.12], respectively).

CONCLUSIONS

The development of Fib4 ≥ 2.67 after 10 years was associated with heavy alcohol drinking and GGT level ≥ 200 IU/L. Therefore, Fib4 combined with GGT could indicate high risk of ALD. However, clinical examinations and course observations are essentially needed.