Loading...
Publication
Influência do Ângulo Foveal na Deteção do Glaucoma em Retinografias
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 14.71 MB | Adobe PDF |
Advisor(s)
Abstract(s)
Introdução O glaucoma é uma das principais causas de cegueira irreversível em todo o mundo, sendo uma doença ocular crónica e progressiva. Recentemente vários estudos avaliaram o impacto do ângulo foveal, também designado por ângulo fóvea-disco (DFA) na deteção do glaucoma em OCT. No entanto, não existem estudos que analisem a influência do DFA na deteção do glaucoma em retinografias. Objetivo Este trabalho pretende avaliar a influência do ângulo foveal na deteção do glaucoma em imagens do fundo do olho, utilizando métricas específicas como o rácio escavação-disco ótico vertical (vCDR), regra ISNT (I>S>N>T), bem como suas variantes: regras IST (I>S>T), IS (I>S), e T (em que a borda temporal do anel neurorretiniano é a menor de todas as 4 bordas). Materiais e Métodos O estudo realizado é de natureza observacional transversal. Para sua realização, foram analisadas 1670 retinografias obtidas de duas bases de dados públicas: ORIGA e G1020. O ângulo foveal (DFA) é definido pela interseção do eixo horizontal que passa pelo centro do disco ótico e o eixo que une a fóvea e essa posição central do disco ótico. Foi considerado dois tipos de DFA. O primeiro é relativo ao DFA original de cada retinografia, designado por ß. Neste ângulo, as medidas são feitas sem a correção do DFA. O segundo é designado por a, com variação angular entre -30º e 20º em passos de 2º, onde 0º representa a medida com a correção exata do DFA. Foi desenvolvido um algoritmo em MATLAB para o cálculo automatizado dos DFA e das métricas vCDR, ISNT e suas variantes, bem como para a criação de gráficos de histogramas, sensibilidade, especificidade, exatidão, F1-Score, curvas ROC com os valores de AUROC e outras análises estatísticas adicionais. Resultados Das 1670 retinografias, 907 (699 normais e 208 glaucomatosos) foram selecionadas e analisadas no presente estudo. O ângulo foveal médio destas imagens foi de -7.6 ± 3.6° (variação: -19.9º a -0.1º). As comparações emparelhadas entre os a e ß não revelaram diferenças estatisticamente significativas para o intervalo angular de a de -12º a 0º (p>0,05) para o vCDR, ISNT e suas variantes. Ao contrário dos resultados obtidos para o vCDR, a regra ISNT e suas variantes apresentaram diferenças nos valores de sensibilidade, especificidade e exatidão entre as diferentes variações angulares de a. No entanto, para as 5 métricas analisadas, os F1-Score dessas métricas não apresentaram variações significativas entre os a, nem entre os a e as medidas realizadas sem correção do DFA (ß). Além disso, as comparações dos valores de AUROC obtidos para cada variação angular de a e ß não apresentaram diferenças estatisticamente significativas (Teste de DeLong, p > 0,1). Conclusão O DFA não influencia a deteção do glaucoma por meio das métricas vCDR, ISNT, IST, IS e T em retinografias. Contudo, a sua utilidade no acompanhamento clínico a longo prazo ainda é uma questão em aberto, o que destaca a importância de estudos adicionais para explorar a importância do DFA nesse contexto.
Introduction Glaucoma is one of the leading causes of irreversible blindness worldwide, being a chronic and progressive ocular disease. Recently, several studies have evaluated the impact of the foveal angle, also referred to as the disc-fovea angle (DFA), on the detection of glaucoma using OCT. However, no studies have analyzed the influence of DFA on glaucoma detection in fundus image. Objective This study aims to evaluate the influence of the foveal angle on glaucoma detection in fundus images, using specific metrics such as the vertical cup-to-disc ratio (vCDR), the ISNT rule (I>S>N>T), as well as its variants: the IST rule (I>S>T), IS (I>S), and T (where the temporal rim of the neuroretinal rim is the thinnest among the four rims). Materials and Methods The study conducted is of a cross-sectional observational nature. A total of 1,670 fundus images from two public databases, ORIGA and G1020, were analyzed. The foveal angle (DFA) is defined as the intersection of the horizontal axis passing through the center of the optic disc and the axis connecting the fovea to the central point of the optic disc. Two types of DFA were considered. The first corresponds to the original DFA of each fundus image, referred to as ß. In this case, measurements were taken without correction of the DFA. The second, referred to as a, exhibits angular variation between -30° and 20° in 2° increments, where 0° represents the measurement with DFA correction. An algorithm was developed in MATLAB for the automated calculation of DFAs and the metrics vCDR, ISNT and its variants. Additionally, the algorithm facilitated the creation of histogram plots, sensitivity, specificity, accuracy, F1-Score, ROC curves with AUROC values, and other supplementary statistical analyses. Results Of the 1670 retinographies, 907 (699 normal and 208 glaucomatous) were selected and analyzed in the present study. The mean foveal angle of these images was -7.6 ± 3.6° (range: -19.9° to - 0.1°). Paired comparisons between a and ß did not reveal statistically significant differences for the angular interval of a from -12° to 0° (p > 0.05) for vCDR, ISNT and its variants.. Contrary to the results obtained for vCDR, the ISNT rule and its variants showed differences in sensitivity, specificity, and accuracy across the different angular variations of a. However, for the 5 metrics analyzed, the F1-scores of these metrics did not show significant variations between the a angles, nor between the a angles and the measurements performed without DFA correction (ß). Additionally, comparisons of the AUROC values obtained for each angular variation of a and ß did not show statistically significant differences (DeLong’s test, p>0.1). Conclusion The DFA does not influence the detection of glaucoma through the vCDR, ISNT, IST, IS, and T metrics in fundus images. However, its utility in long-term clinical monitoring remains an open question, highlighting the importance of further studies to explore the role of DFA in this context.
Introduction Glaucoma is one of the leading causes of irreversible blindness worldwide, being a chronic and progressive ocular disease. Recently, several studies have evaluated the impact of the foveal angle, also referred to as the disc-fovea angle (DFA), on the detection of glaucoma using OCT. However, no studies have analyzed the influence of DFA on glaucoma detection in fundus image. Objective This study aims to evaluate the influence of the foveal angle on glaucoma detection in fundus images, using specific metrics such as the vertical cup-to-disc ratio (vCDR), the ISNT rule (I>S>N>T), as well as its variants: the IST rule (I>S>T), IS (I>S), and T (where the temporal rim of the neuroretinal rim is the thinnest among the four rims). Materials and Methods The study conducted is of a cross-sectional observational nature. A total of 1,670 fundus images from two public databases, ORIGA and G1020, were analyzed. The foveal angle (DFA) is defined as the intersection of the horizontal axis passing through the center of the optic disc and the axis connecting the fovea to the central point of the optic disc. Two types of DFA were considered. The first corresponds to the original DFA of each fundus image, referred to as ß. In this case, measurements were taken without correction of the DFA. The second, referred to as a, exhibits angular variation between -30° and 20° in 2° increments, where 0° represents the measurement with DFA correction. An algorithm was developed in MATLAB for the automated calculation of DFAs and the metrics vCDR, ISNT and its variants. Additionally, the algorithm facilitated the creation of histogram plots, sensitivity, specificity, accuracy, F1-Score, ROC curves with AUROC values, and other supplementary statistical analyses. Results Of the 1670 retinographies, 907 (699 normal and 208 glaucomatous) were selected and analyzed in the present study. The mean foveal angle of these images was -7.6 ± 3.6° (range: -19.9° to - 0.1°). Paired comparisons between a and ß did not reveal statistically significant differences for the angular interval of a from -12° to 0° (p > 0.05) for vCDR, ISNT and its variants.. Contrary to the results obtained for vCDR, the ISNT rule and its variants showed differences in sensitivity, specificity, and accuracy across the different angular variations of a. However, for the 5 metrics analyzed, the F1-scores of these metrics did not show significant variations between the a angles, nor between the a angles and the measurements performed without DFA correction (ß). Additionally, comparisons of the AUROC values obtained for each angular variation of a and ß did not show statistically significant differences (DeLong’s test, p>0.1). Conclusion The DFA does not influence the detection of glaucoma through the vCDR, ISNT, IST, IS, and T metrics in fundus images. However, its utility in long-term clinical monitoring remains an open question, highlighting the importance of further studies to explore the role of DFA in this context.
Description
Keywords
Ângulo Foveal Cabeça do Nervo Ótico Glaucoma Rácio Escavação-Disco Ótico Regra Isnt e Variantes Retinografia