Why carry out this study? |
Artificial intelligence can enhance our ability to diagnose eye diseases like keratoconus by leveraging large volumes of ocular images and training models with deep learning algorithms. |
The significant change in corneal endothelial cells observed in the diagnosis of keratoconus has not been significantly explored in previous studies. |
What was learned from this study? |
Our research developed an end-to-end model to automatically identify and assess corneal endothelial morphological changes in keratoconus eyes. |
We also constructed a novel nomogram, which can provide valuable information for the diagnosis, monitoring, and management of the disease. |
Introduction
-
Mechanical stress. The progressive thinning and steepening of the cornea in keratoconus result in increased mechanical stress on the corneal endothelium. This mechanical stress can potentially lead to endothelial cell damage or dysfunction;
-
Oxidative stress. Keratoconus is associated with increased levels of oxidative stress in the cornea. Reactive oxygen species (ROS), which are harmful molecules generated during various metabolic processes, can damage the corneal endothelial cells and impair their function.
-
Inflammatory factors. Chronic inflammation has been implicated in the pathogenesis of keratoconus. Inflammatory mediators released in the cornea can affect the corneal endothelial cells and contribute to endothelial changes.
Methods
Study Design and Datasets
Specular Microscopy
Corneal Endothelial Segmentation
Keratoconus Classification
Statistical Analysis
Results
Participants and Datasets
Characteristic | N = 403 eyes (221 patients) |
---|---|
Sex, n (%) | |
Female | 68 (31%) |
Male | 153 (69%) |
Age, years | 22 [18, 27] |
Sphere, D | − 4.2 [− 7.0, − 1.7)] |
Cylinder, D | − 2.00 [− 4.00, − 1.00)] |
Axis, D | 90 [35, 146] |
CDVA (logMAR) | 0.19 [0.03, 0.40] |
K1, D | 45.2 [43.2, 48.9 |
K2, D | 49 [46, 53] |
Stage, n (%) | |
I | 93 (23%) |
II | 90 (22%) |
III&IV | 220 (55%) |
CCT, μm | 466 [439, 496] |
ECD, cells/mm2 | 2988 [2798, 3180] |
CV | 30 [27, 33] |
HEX, % | 57 [48, 66] |
N, count | 175 [113, 240] |
MIN | 149 [137, 160] |
MAX | 666 [614, 749] |
AVG | 335 [315, 357] |
SD | 99 [89, 110] |
Level | Normal | Keratoconus | p | Statistical test |
---|---|---|---|---|
Number of eyes | 370 | 403 | ||
Sex, n (%) | ||||
Female | 142 (76.8%) | 68 (31%) | 0.00 | χ2 |
Male | 43(23.2%) | 153 (69%) | ||
Age, years | 25.00 [22.00, 30.00] | 22.00 [18.00, 27.00] | 0.00 | Wilcox |
SE, D | − 11.50 [− 13.00, − 9.00] | − 5.50 [− 8.50, − 3.00] | 0.00 | Wilcox |
K1, D | 43.42 [42.18, 44.32] | 45.20 [43.20, 48.90] | 0.00 | Wilcox |
K2, D | 44.66 [43.64, 45.79] | 48.80 [45.60, 52.82] | 0.00 | Wilcox |
CCT, μm | 505.50 [488.00, 529.75] | 466.00 [439.00, 496.00] | 0.00 | Wilcox |
ECD, cells/mm2 | 3014.00 [2828.25, 3192.75] | 2988.00 [2798.50, 3180.00] | 0.20 | Wilcox |
CV | 31.00 [28.00, 34.00] | 30.00 [27.00, 33.00] | 0.00 | Wilcox |
HEX, % | 55.00 [49.00, 61.00] | 57.00 [48.00, 66.00] | 0.02 | Wilcox |
N, count | 255.50 [213.00, 285.75] | 175.00 [113.00, 240.50] | 0.00 | Wilcox |
MIN | 143.00 [132.00, 152.00] | 149.00 [137.00, 160.00] | 0.00 | Wilcox |
MAX | 700.00 [641.00, 773.75] | 666.00 [614.50, 749.00] | 0.00 | Wilcox |
AVG | 332.00 [313.25, 353.75] | 335.00 [315.00, 357.00] | 0.20 | Wilcox |
SD | 102.00 [94.00, 116.00] | 99.00 [89.25, 110.00] | 0.00 | Wilcox |
DL features | ||||
---|---|---|---|---|
IN-0 | − 0.11 [− 0.24, 0.12] | − 0.16 [− 0.68, 0.52] | 0.01 | Wilcox |
IN-1 | 0.01 [− 0.08, 0.08] | 0.00 [− 0.12, 0.11] | 0.74 | Wilcox |
RS-0 | − 0.31 [− 0.68, 0.48] | − 0.24 [− 1.43, 1.15] | 0.18 | Wilcox |
RS-1 | 0.00 [− 0.37, 0.30] | − 0.09 [− 0.48, 0.48] | 0.53 | wilcox |
DS-0 | − 3.93 [− 5.35, − 0.51] | − 19.34 [− 31.63, 18.83] | 0.00 | Wilcox |
DS-1 | − 0.49 [− 2.15, 1.62] | 1.25 [− 6.05, 4.16] | 0.02 | Wilcox |
MB-0 | − 0.08 [− 0.31, 0.20] | − 0.42 [− 0.87, 0.73] | 0.00 | Wilcox |
MB-1 | − 0.02 [− 0.23, 0.22] | − 0.06 [− 0.40, 0.27] | 0.10 | Wilcox |
Variables | Training dataset | Test dataset | Validation dataset | p | Statistical test |
---|---|---|---|---|---|
Number of eyes | 273 | 116 | 383 | ||
Sex, n (%) | |||||
Female | 84 (54.2%) | 27 (45.7%) | 99 (51.5%) | 0.47 | χ2 |
Male | 71 (45.8%) | 32 (54.3%) | 93 (48.5%) | ||
Age | 24.00 [20.00, 29.00] | 24.00 [20.00, 27.00] | 24.00 [20.00, 29.00] | 0.48 | Kruskal–Wallis |
SE | − 8.50 [− 12.00, − 4.75] | − 8.62 [− 11.50, − 4.91] | − 8.75 [− 12.00, − 5.50] | 0.63 | Kruskal–Wallis |
K1 | 43.82 [42.52, 45.40] | 44.10 [42.80, 45.91] | 43.93 [42.60, 45.90] | 0.29 | Kruskal–Wallis |
K2 | 45.73 [44.02, 48.50] | 46.10 [44.36, 49.98] | 45.75 [44.09, 49.20] | 0.49 | Kruskal–Wallis |
CCT | 491.00 [462.00, 515.00] | 487.50 [454.00, 508.25] | 491.00 [460.50, 517.00] | 0.40 | Kruskal–Wallis |
ECD | 3009.00 [2815.00, 3199.00] | 2980.50 [2772.25, 3200.00] | 2999.00 [2824.00, 3176.00] | 0.71 | Kruskal–Wallis |
CV | 31.00 [28.00, 33.00] | 30.00 [27.00, 34.00] | 30.00 [28.00, 33.00] | 0.22 | Kruskal–Wallis |
HEX | 56.00 [48.00, 62.00] | 56.00 [46.00, 63.00] | 56.00 [51.00, 64.00] | 0.26 | Kruskal–Wallis |
N | 220.00 [160.00, 275.00] | 224.00 [154.25, 259.25] | 218.00 [153.50, 267.00] | 0.63 | Kruskal–Wallis |
MIN | 145.00 [136.00, 156.00] | 147.00 [132.75, 158.25] | 146.00 [135.00, 156.00] | 0.96 | Kruskal–Wallis |
MAX | 679.00 [620.00, 767.00] | 695.50 [639.25, 772.25] | 679.00 [627.00, 755.00] | 0.36 | Kruskal–Wallis |
AVG | 332.00 [313.00, 355.00] | 335.50 [312.50, 361.00] | 333.00 [315.00, 354.00] | 0.72 | Kruskal–Wallis |
SD | 101.00 [93.00, 112.00] | 102.00 [92.75, 111.50] | 100.00 [91.00, 113.75] | 0.45 | Kruskal–Wallis |