Machine learning classifiers for glaucoma diagnosis based on classification of retinal nerve fibre layer thickness parameters measured by Stratus OCT

Purpose: To compare the performance of two machine learning classifiers (MLCs), artificial neural networks (ANNs) and support vector machines (SVMs), with input based on retinal nerve fibre layer thickness (RNFLT) measurements by optical coherence tomography (OCT), on the diagnosis of glaucoma, and to assess the effects of different input parameters. Methods: We analysed Stratus OCT data from 90 healthy persons and 62 glaucoma patients. Performance of MLCs was compared using conventional OCT RNFLT parameters plus novel parameters such as minimum RNFLT values, 10th and 90th percentiles of measured RNFLT, and transformations of A‐scan measurements. For each input parameter and MLC, the area under the receiver operating characteristic curve (AROC) was calculated. Results: There were no statistically significant differences between ANNs and SVMs. The best AROCs for both ANN (0.982, 95%CI: 0.966–0.999) and SVM (0.989, 95% CI: 0.979–1.0) were based on input of transformed A‐scan measurements. Our SVM trained on this input performed better than ANNs or SVMs trained on any of the single RNFLT parameters (p ≤ 0.038). The performance of ANNs and SVMs trained on minimum thickness values and the 10th and 90th percentiles were at least as good as ANNs and SVMs with input based on the conventional RNFLT parameters. Conclusion: No differences between ANN and SVM were observed in this study. Both MLCs performed very well, with similar diagnostic performance. Input parameters have a larger impact on diagnostic performance than the type of machine classifier. Our results suggest that parameters based on transformed A‐scan thickness measurements of the RNFL processed by machine classifiers can improve OCT‐based glaucoma diagnosis.     

Estimating the rate of progressive visual field damage in those with open-angle glaucoma, from cross-sectional data

PURPOSE: To estimate the rate of visual field progression in open-angle glaucoma (OAG) subjects, by using data from population-based cross-sectional studies. METHODS: Subjects with OAG were identified in nine surveys of randomly sampled populations using standard criteria for glaucomatous optic neuropathy. Subjects were of European, African, Chinese, and Hispanic ethnicity. The measure of OAG damage was the mean deviation (MD) of an automated visual field test (Humphrey Field Analyzer; Carl Zeiss Meditec, Inc., Dublin, CA). The rate of progression was the mean of all subjects' damage in the worse eye divided by an average time since onset. Time since onset was estimated from age-specific prevalence rates. RESULTS: A total of 1066 subjects with OAG contributed visual field data. The mean worsening in decibels per year was: European-derived, -1.12; Hispanic, -1.26; African-derived, -1.33; and Chinese -1.56 (difference among ethnicities, P = 0.16). The mean duration of disease was lowest among Chinese persons at 10.5 years (95% CI: 8.8-12.6) and was highest in African-derived subjects at 15.4 years (95% CI: 14.6-15.9). The progression rate was not consistently related to age or gender. By combining disease duration and progression rate, the model predicted that 15% or fewer of the worse eyes would reach the end of the field damage scale in the patient's lifetime. CONCLUSIONS: The estimates of typical worsening per year in the worse eye among subjects with OAG suggested slightly more rapid progression than in some clinic-based studies. The rate did not differ significantly by ethnicity or gender, but was worse in those with known, treated OAG and in pseudophakic subjects.     

Intraocular pressure reduction with a fixed treatment protocol in the Early Manifest Glaucoma Trial

Purpose: To evaluate: (i) the relationship between intraocular pressure (IOP) reduction attained with a fixed treatment protocol and the untreated IOP level; (ii) the consistency of IOP reduction over time; and (iii) whether there is a threshold pretreatment IOP level below which IOP reduction might be less effective. Results are based on 128 patients with glaucoma with field defects, who were randomized to the treatment arm of the Early Manifest Glaucoma Trial (EMGT). Methods: The EMGT fixed treatment protocol consisted of 360° laser trabeculoplasty and topical betaxolol eye drops B.I.D. Treatment was unchanged as long as progression did not occur. Analyses assessed the initial IOP reduction after 3 months and also the mean reduction based on all follow‐up values; IOP changes over time were evaluated with linear regression analysis. Factors influencing initial and mean IOP reduction were also explored using linear models. Results: Mean age at baseline was 68 years, and untreated baseline IOP ranged from 13 to 30.5 mmHg. On average, eyes with higher baseline IOP experienced larger pressure reductions than eyes with lower baseline IOP, whether expressed in mmHg or as percentages. Each mmHg of higher baseline IOP was associated with approximately 0.6 mmHg larger IOP reduction. IOP changed little over time, with 66% of patients changing less than 0.5 mmHg/year, and only 13% (17/128) changing >1.0 mmHg/year. The treatment protocol did not achieve any average IOP reduction in eyes with baseline pressures ≤ 15 mmHg. Factors related to more IOP reduction at 3 months were higher baseline IOP and positive refractive error, while higher baseline IOP and male gender (more reduction) and cardiovascular disease history (less reduction) were associated with mean IOP on treatment. Conclusion: With a fixed treatment protocol, the IOP reduction achieved depended very strongly on baseline untreated IOP levels. There seemed to be a lower threshold around 15 mmHg, where therapy did not result in any reduction of IOP. Our results suggest that when effects of IOP‐lowering treatment are reported, whether expressed in mmHg or as a percentage of untreated pressure levels, the baseline IOP levels should be specified as well.     

A novel mitochondrial transfer RNA(Asn) mutation causing multiorgan failure

BACKGROUND: Mitochondrial cytopathies are a heterogeneous group of disorders with a broad spectrum of clinical symptoms. OBJECTIVE: To characterize a novel mutation in the transfer RNA(Asn) (m.5728A>G) identified in a 13-year-old boy with multiorgan failure. DESIGN: Biochemical and immunocytochemical studies were performed in combination with transmitochondrial cybrid analysis. SETTING: A university hospital. Molecular and biochemical analyses were performed in collaboration between 2 other university hospitals. PATIENT: Thirteen-year-old boy with multiorgan failure. RESULTS: In the patient's muscle tissue and cultured skin fibroblasts, a combined deficiency of complexes I and IV was found, using spectrophotometric analysis and activity staining in the gel following blue native polyacrylamide gel electrophoresis. An identical biochemical profile was seen in transmitochondrial cybrids carrying more than 55% mutant mitochondrial DNA. CONCLUSION: These data suggest that the m.5728A>G transition is a pathogenic mutation and is the cause of the respiratory chain dysfunction in the propositus.     

Trained artificial neural network for glaucoma diagnosis using visual field data: a comparison with conventional algorithms

PURPOSE: To evaluate and confirm the performance of an artificial neural network (ANN) trained to recognize glaucomatous visual field defects, and compare its diagnostic accuracy with that of other algorithms proposed for the detection of visual field loss. METHODS: SITA Standard 30-2 visual fields, from 100 glaucoma patients and 116 healthy participants, formed the data set. Our ANN was a previously described fully trained network using scored pattern deviation probability maps as input data. Its diagnostic accuracy was compared to that of the Glaucoma Hemifield Test, the Pattern Standard Deviation index at the P<5% and <1%, and also to a technique based on the recognizing clusters of significantly depressed test points. RESULTS: The included tests had early to moderate visual field loss (median MD=-6.16 dB). ANN achieved a sensitivity of 93% at a specificity level of 94% with an area under the receiver operating characteristic curve of 0.984. Glaucoma Hemifield Test attained a sensitivity of 92% at 91% specificity. Pattern Standard Deviation, with a cut off level at P<5% had a sensitivity of 89% with a specificity of 93%, whereas at P<1% the sensitivity and specificity was 72% and 97%, respectively. The cluster algorithm yielded a sensitivity of 95% and a specificity of 82%. CONCLUSIONS: The high diagnostic performance of our ANN based on refined input visual field data was confirmed in this independent sample. Its diagnostic accuracy was slightly to considerably better than that of the compared algorithms. The results indicate the large potential for ANN as an important clinical glaucoma diagnostic tool.     

A microsurgical method for experimental kidney denervation

A successful microsurgical method for denervation of the kidney in the rat is presented. Unilateral kidney denervation was performed in eight animals. Ten sham-operated animals were used as controls. Under ether anaesthesia, the kidney vessels on one side were exposed. The peritoneum above the vessels was incised and the area was stained with 1% toluidine blue solution. The nerves were identified. With microsurgical technique, the nerves were carefully separated from the vessels and resected. Tissue specimens from both kidneys in all animals were removed and analysed for norepinephrine content. There was no normal difference between the kidneys in the sham-operated rats. About 1 week after denervation, the norepinephrine concentration in the denervated kidney showed a 95% reduction, indicating a complete denervation. The described microsurgical procedure was a rapid, simple, and reproducible method for kidney denervation in the rat.     

Membranous duodenal stenosis

The experience of our 16 patients treated for membranous duodenal stenosis is reported. Their treatment and course was analysed in a retrospective study. Eight patients were operated on within the first 16 days of life and in the remaining group surgery was performed at 1 month to 4 y of age. The presenting symptom leading to diagnosis was, in all but one case, non-bile-stained vomiting. Associated malformations were found in all but four patients, mostly morbus Down. The operative procedure performed was a partial excision of the duodenal membrane and a duodenoplasty in 10 patients, a duodenojejunostomy in 5 patients, and a duodenoplasty only in 1 patient. The postoperative course was without lethal complications. One late stricture in an anastomosis occurred. We conclude that in its presentation, duodenal stenosis differs from duodenal atresia, and can often be misinterpreted, resulting in a late diagnosis, and should be reported as a separate entity.