Deep learning is being employed in disease detection and classification based on medical images for clinical decision making. It typically requires large amounts of labelled data; however, the sample ...size of such medical image datasets is generally small. This study proposes a novel training framework for building deep learning models of disease detection and classification with small datasets. Our approach is based on a hierarchical classification method where the healthy/disease information from the first model is effectively utilized to build subsequent models for classifying the disease into its sub-types via a transfer learning method. To improve accuracy, multiple input datasets were used, and a stacking ensembled method was employed for final classification. To demonstrate the method's performance, a labelled dataset extracted from volumetric ophthalmic optical coherence tomography data for 156 healthy and 798 glaucoma eyes was used, in which glaucoma eyes were further labelled into four sub-types. The average weighted accuracy and Cohen's kappa for three randomized test datasets were 0.839 and 0.809, respectively. Our approach outperformed the flat classification method by 9.7% using smaller training datasets. The results suggest that the framework can perform accurate classification with a small number of medical images.
Full text
Available for:
IZUM, KILJ, NUK, PILJ, PNG, SAZU, UL, UM, UPUK
Purpose
To determine whether choroidal blood flow (BF) is related to visual field (VF) defect severity and progression in eyes with open-angle glaucoma (OAG).
Study design
Retrospective and ...longitudinal.
Methods
This study comprised 443 eyes of 285 OAG patients who underwent laser speckle flowgraphy (LSFG), optical coherence tomography, and visual-field (VF) testing at baseline. The patients were then observed for at least 2 years and at least 5 reliable VF tests were performed. In the LSFG images, we set regions of interest at the optic nerve head (ONH) and the parapapillary choroid to obtain ONH-tissue mean blur rate (MBR) and choroidal MBR, respectively. We used univariable and multivariable linear mixed-effects models to determine clinical factors associated with choroidal MBR at baseline. We also used a linear mixed-effects model to determine the contribution of ONH-tissue MBR and choroidal MBR to baseline mean deviation (MD) and to MD slope during follow-up, adjusting for potential confounding factors, including circumpapillary retinal nerve fiber layer thickness.
Results
Choroidal MBR was associated with age, MD slope, and ONH-tissue MBR (β = -0.181,
P
= 0.001; β = 0.134,
P
= 0.002; β = 0.096,
P
= 0.049, respectively). ONH-tissue MBR was associated with both MD and MD slope (β = 0.146,
P
= 0.004; β = 0.152,
P
= 0.009, respectively), whereas choroidal MBR was associated only with MD slope (β = 0.147,
P
= 0.005).
Conclusion
LSFG-derived choroidal MBR might be a useful biomarker to predict VF defect progression in a Japanese population.
Full text
Available for:
EMUNI, FIS, FZAB, GEOZS, GIS, IJS, IMTLJ, KILJ, KISLJ, MFDPS, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, SBMB, SBNM, UKNU, UL, UM, UPUK, VKSCE, ZAGLJ
To investigate the effect of K-115, a novel Rho kinase (ROCK) inhibitor, on retinal ganglion cell (RGC) survival in an optic nerve crush (NC) model. Additionally, to determine the details of the ...mechanism of K-115's neuroprotective effect in vivo and in vitro.
ROCK inhibitors, including K-115 and fasudil (1 mg/kg/d), or vehicle were administered orally to C57BL/6 mice. Retinal ganglion cell death was then induced with NC. Retinal ganglion cell survival was evaluated by counting surviving retrogradely labeled cells and measuring RGC marker expression with quantitative real-time polymerase chain reaction (qRT-PCR). Total oxidized lipid levels were assessed with a thiobarbituric acid-reactive substances (TBARS) assay. Reactive oxygen species (ROS) levels were assessed by co-labeling with CellROX and Fluorogold. Expression of the NADPH oxidase (Nox) family of genes was evaluated with qRT-PCR.
The survival of RGCs after NC was increased 34 ± 3% with K-115, a significantly protective effect. Moreover, a similar effect was revealed by the qRT-PCR analysis of Thy-1.2 and Brn3a, RGC markers. Levels of oxidized lipids and ROS also increased with time after NC. NC-induced oxidative stress, including oxidation of lipids and production of ROS, was significantly attenuated by K-115. Furthermore, expression of the Nox gene family, especially Nox1, which is involved in the NC-induced ROS production pathway, was dramatically reduced by K-115.
The results indicated that oral K-115 administration delayed RGC death. Although K-115 may be mediated through Nox1 downregulation, we found that it did not suppress ROS production directly. Our findings show that K-115 has a potential use in neuroprotective treatment for glaucoma and other neurodegenerative diseases.
Detection, diagnosis, and treatment of ophthalmic diseases depend on extraction of information (features and/or their dimensions) from the images. Deep learning (DL) model are crucial for the ...automation of it. Here, we report on the development of a lightweight DL model, which can precisely segment/detect the required features automatically. The model utilizes dimensionality reduction of image to extract important features, and channel contraction to allow only the required high-level features necessary for reconstruction of segmented feature image. Performance of present model in detection of glaucoma from optical coherence tomography angiography (OCTA) images of retina is high (area under the receiver-operator characteristic curve AUC ~ 0.81). Bland-Altman analysis gave exceptionally low bias (~ 0.00185), and high Pearson's correlation coefficient (p = 0.9969) between the parameters determined from manual and DL based segmentation. On the same dataset, bias is an order of magnitude higher (~ 0.0694, p = 0.8534) for commercial software. Present model is 10 times lighter than Unet (popular for biomedical image segmentation) and have a better segmentation accuracy and model training reproducibility (based on the analysis of 3670 OCTA images). High dice similarity coefficient (D) for variety of ophthalmic images suggested it's wider scope in precise segmentation of images even from other fields. Our concept of channel narrowing is not only important for the segmentation problems, but it can also reduce number of parameters significantly in object classification models. Enhanced disease diagnostic accuracy can be achieved for the resource limited devices (such as mobile phone, Nvidia's Jetson, Raspberry pi) used in self-monitoring, and tele-screening (memory size of trained model ~ 35 MB).
Full text
Available for:
IZUM, KILJ, NUK, PILJ, PNG, SAZU, UL, UM, UPUK
To investigate waveform changes in blood flow (BF) in the optic nerve head (ONH) and to evaluate their usefulness in identifying normal tension glaucoma (NTG).
Sixty-one eyes of 61 patients with NTG ...and 21 eyes of age-matched healthy control subjects were included in this study. The NTG eyes were divided into the following three groups based on the progression of their visual field defects: mild (mean deviation MD greater than -6.0 decibels dB), moderate (MD between -6.0 and -12.0 dB), and severe (MD less than -12.0 dB). The ONH BF analysis was performed with laser speckle flowgraphy (LSFG) and included waveform variables such as skew, acceleration time index (ATI), and blowout time.
In the ONH, LSFG skew variables were significantly lower in the NTG eyes than in the control eyes (P < 0.001), and ATI was significantly higher (P < 0.01), despite similar systemic characteristics in the four groups. The differences were most marked in the mild NTG group. Multiple linear regression analysis showed that MD, average thickness of the circumpapillary retinal nerve fiber layer, and pulse rate were predictive factors for both skew and ATI. A receiver operating characteristic (ROC) curve analysis also revealed that skew (area under the ROC curve, 0.89) and ATI (area under the ROC curve, 0.80) had the greatest power to differentiate normal eyes from eyes with mild NTG.
These results suggest that LSFG measurements of waveform changes in ONH BF can differentiate healthy eyes from eyes with NTG, particularly those with mild NTG.
To investigate the effect of mitochondrial dysfunction on the autoregulation of blood flow, by measuring levels of glutathione, an indicator of mitochondrial dysfunction, in glaucoma patients.
...Fifty-six OAG patients and 21 age-matched controls underwent a blood assay. Mitochondrial function was measured according to the levels of total glutathione (t-GSH), reduced GSH (GSH), and oxidized GSH (GSSG, glutathione disulfide) in peripheral blood mononuclear cells. Ocular blood flow in the optic nerve head was assessed with laser speckle flowgraphy parameters, including acceleration time index (ATI). We determined correlations between these measurements and other clinical parameters. Furthermore, we investigated the association between glutathione levels and glaucoma with a logistic regression analysis. Finally, we calculated the area under the receiver operating characteristic (ROC) curve in order to determine the power of redox index (the log GSH/GSSG ratio) to distinguish the groups.
OAG patients demonstrated significantly higher GSSG levels and a lower redox index than the controls (p = 0.01, p = 0.01, respectively), but total GSH and reduced GSH levels were similar in the OAG subjects and controls (p = 0.80, p = 0.94, respectively). Additionally, redox index was significantly correlated with mean deviation (MD) of the visual field (r = 0.29, p = 0.03) and ATI (r = -0.30, p = 0.03). Multiple linear regression analysis showed that redox index contributed to MD (p = 0.02) and ATI (p = 0.04). The receiver operating characteristic curve (AUC) analysis suggested that redox index could differentiate between control eyes and eyes with glaucoma (AUC; 0.70: 95% interval; 0.57-0.84). The cutoff point for redox index to maximize its sensitivity and specificity was 2.0 (sensitivity: 91.1%, specificity: 42.9%).
These results suggest that redox index is lower in OAG patients than in controls. Thus, it is possible that mitochondrial dysfunction contributes to glaucoma pathogenesis by causing vascular alterations.
Full text
Available for:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
To investigate sectoral differences in the relationship between optic nerve head tissue blood flow, anatomically corresponding visual field defect severity, and future progression.
This retrospective ...longitudinal medical chart review comprised 508 eyes of 319 open-angle glaucoma patients (mean deviation: -9.2 ± 7.0 dB), followed for an average of 4.7 ± 1.1 years; an average 11.7 ± 3.7 visual field tests were performed. Average total deviation (TD) was calculated in the superior, central, and inferior sectors of the Humphrey 24-2 program. The optic nerve head was divided to obtain inferior, temporal, and superior tissue-area mean blur rate (MT), derived from laser speckle flowgraphy. At baseline, the correlation between MT and TD was compared in anatomically corresponding sectors. We performed a multivariate analysis to determine the contribution of baseline MT to corresponding TD slope and to determine background factors influencing superior to temporal MT. We used a linear-mixed effect model for the statistical analysis.
At baseline, the highest β coefficients were found between MT-superior and TD-inferior, MT-temporal and TD-central, and between MT-inferior and TD-superior, in that order (β = 0.38, β = 0.27, β = 0.26, respectively). MT-superior and MT-temporal independently contributed to corresponding TD slope (P < 0.05). Male sex, high body mass index, and the prevalence of sleep apnea syndrome were contributing factors to lower superior to temporal MT (P < 0.05).
Review of medical history, measurements of systemic variables, and laser speckle flowgraphy parameters might help clinicians to predict visual field defect severity and progression.
To investigate the relationship between various characteristics of a normal population and laser speckle flowgraphy (LSFG) measurements of mean blur rate (MBR) in the optic nerve head (ONH).
A total ...of 189 eyes of 189 normal subjects (93 male, 96 female, mean age 45 ± 14 years old, age range: 20-72) without any history of hypertension, hyperlipidemia or diabetes were enrolled. ONH microcirculation was measured with LSFG and overall MBR (MA), vessel-area MBR (MV), and tissue-area MBR (MT) were derived from these measurements. The statistical association of these measurements with characteristics such as sex, age, intraocular pressure (IOP) and systolic blood pressure (SBP) was then determined.
There was a trend towards decreased IOP and MV and increased SBP with age (P = 0.002, P = 0.035, and P = 0.006, respectively). Furthermore, IOP, MV and SBP were correlated with age (r = -0.23, P = 0.011; r = -0.24, P < 0.001; and r = 0.30, P < 0.001, respectively). Separate multiple regression analyses of independent contributing factors revealed that sex and IOP contributed to MA (P < 0.001 and P = 0.002, respectively), sex, IOP, and age contributed to MV (P < 0.001, P = 0.003, and P = 0.024, respectively), while only IOP contributed to MT (P = 0.003).
In a normal population, MBR was affected by IOP in both the large vessel and capillary areas of the ONH, but not by SBP. MV was also affected by age and sex, while MT was stable independent of age or sex.
Full text
Available for:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
Systemic antioxidative status has been implicated in glaucoma pathogenesis. Additionally, corneal hysteresis (CH) may contribute to glaucoma progression. Here, we evaluated the relationship between ...biological antioxidant potential (BAP) and CH. This study included 103 patients with open-angle glaucoma (OAG). We used a free radical analyzer to measure BAP, and an ocular response analyzer to measure CH and corneal resistance factor (CRF). We evaluated the relationship between systemic oxidative stress and other clinical parameters with Spearman's rank correlation test and a multi-regression analysis. BAP was not correlated to either CH or CRF in the male or female OAG patients. BAP was correlated to both CH and CRF in the female OAG patients older than 57 years (r = 0.51, P = 0.003; r = 0.49, P = 0.004), but uncorrelated in the female OAG patients younger than 57 years. Multiple regression analysis revealed that BAP independently contributed to CH (P = 0.025) and CRF (P = 0.015) in the older female OAG patients. Systemic oxidative stress may significantly affect the viscoelasticity of the cornea in older female OAG patients. Future studies are needed to confirm that low systemic antioxidative status and low corneal hysteresis contribute to glaucoma pathogenesis.
Full text
Available for:
IZUM, KILJ, NUK, PILJ, PNG, SAZU, UL, UM, UPUK
PDF
