Diagnostic Quality Assessment of Ocular Fundus Photographs: Efficacy of Structure-Preserving ScatNet Features

Dev, Chander; M.S, Sharang; Manne, Shanmukh Reddy; Goud, Abhilash; Bashar, Sarforaz Bin; Richhariya, Ashutosh; Chhablani, Jay; Vupparaboina, Kiran Kumar; Jana, Soumya

doi:10.1109/embc.2019.8857046

Cited by 3 publications

(2 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The performance indices obtained from various models are presented in Table . 1. Clearly, the proposed DL-based models: ResNet18, ENet-B0, and ENet-B3 perform significantly better than previously reported natural IQA-metric-based methods including BRISQUE 40 , NBIQA 41 , and ScatNets 26 . In particular, among natural IQA-metric-based methods, BRISQUE and NBIQA, respectively, achieved mean accuracy values of 62.73% and 71.14% which is relatively poor vis-à-vis corresponding accuracy value 87.54% obtained by ScanNet based approach.…”

Section: Oct Choroid Quality Assessmentmentioning

confidence: 66%

“…More specifically, majority of attempts were directed towards DQA of fundus photography (FP) images focusing on accurate detection of specific diseases such as diabetic retinopathy (DR) 25 . Further, DQA of FPs has been addressed using traditional features 26 , wavelet-based deep scattering features 27 and deep learning (DL)-based methods 28,29 . On the other hand, DQA of OCT images is relatively less explored.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning Based Diagnostic Quality Assessment of Choroidal OCT Features with Expert-Evaluated Explainability

Koidala

Manne

Ozimba

et al. 2022

Preprint

View full text Add to dashboard Cite

Various vision-threatening eye diseases including age-related macular degeneration (AMD) and central serous chorioretinopathy(CSCR) are caused due to the dysfunctions manifested in the highly vascular choroid layer of the posterior segment of the eye.In the current clinical practice, screening choroidal structural changes is widely based on optical coherence tomography (OCT)images. Accordingly, to assist clinicians, several automated choroidal biomarker detection methods using OCT images aredeveloped. However, the performance of these algorithms is largely constrained by the quality of the OCT scan. Consequently,determining the quality of choroidal features in OCT scans is significant in building standardized quantification tools and henceconstitutes our main objective. This study includes a dataset of 1593 good and 2581 bad quality Spectralis OCT images gradedby an expert. Noting the efficacy of deep-learning (DL) in medical image analysis, we propose to train three state-of-the-artDL models: ResNet18, EfficientNet-B0 and EfficientNet-B3 to detect the quality of OCT images. The choice of these modelswas inspired by their ability to preserve the salient features across all the layers without information loss. To evaluate theattention of DL models on the choroid, we introduced color transparency maps (CTMs) based on GradCAM explanations.Further, we proposed two subjective grading scores: overall choroid coverage (OCC) and choroid coverage in the visibleregion(CCVR) based on CTMs to objectively correlate visual explanations vis-à-vis DL model attentions. We observed that theaverage accuracy and F-scores for the three DL models are greater than 96%. Further, the OCC and CCVR scores achievedfor the three DL models under consideration substantiate that they mostly focus on the choroid layer in making the decision. Inparticular, of the three DL models, EfficientNet-B3 is in close agreement with the clinician’s inference. The proposed DL-basedframework demonstrated high detection accuracy as well as attention on the choroid layer, where EfficientNet-B3 reportedsuperior performance. Our work assumes significance in bench-marking the automated choroid biomarker detection tools andfacilitating high-throughput screening. Further, the methods proposed in this work can be adopted for evaluating the attentionof DL-based approaches developed for other region-specific quality assessment tasks.

show abstract

Section: Oct Choroid Quality Assessmentmentioning

confidence: 66%

Section: Introductionmentioning

confidence: 99%

Deep Learning Based Diagnostic Quality Assessment of Choroidal OCT Features with Expert-Evaluated Explainability

Koidala

Manne

Ozimba

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

Deep learning based diagnostic quality assessment of choroidal OCT features with expert-evaluated explainability

Koidala

Manne

Ozimba

et al. 2023

Sci Rep

View full text Add to dashboard Cite

Various vision-threatening eye diseases including age-related macular degeneration (AMD) and central serous chorioretinopathy (CSCR) are caused due to the dysfunctions manifested in the highly vascular choroid layer of the posterior segment of the eye. In the current clinical practice, screening choroidal structural changes is widely based on optical coherence tomography (OCT) images. Accordingly, to assist clinicians, several automated choroidal biomarker detection methods using OCT images are developed. However, the performance of these algorithms is largely constrained by the quality of the OCT scan. Consequently, determining the quality of choroidal features in OCT scans is significant in building standardized quantification tools and hence constitutes our main objective. This study includes a dataset of 1593 good and 2581 bad quality Spectralis OCT images graded by an expert. Noting the efficacy of deep-learning (DL) in medical image analysis, we propose to train three state-of-the-art DL models: ResNet18, EfficientNet-B0 and EfficientNet-B3 to detect the quality of OCT images. The choice of these models was inspired by their ability to preserve the salient features across all the layers without information loss. To evaluate the attention of DL models on the choroid, we introduced color transparency maps (CTMs) based on GradCAM explanations. Further, we proposed two subjective grading scores: overall choroid coverage (OCC) and choroid coverage in the visible region(CCVR) based on CTMs to objectively correlate visual explanations vis-à-vis DL model attentions. We observed that the average accuracy and F-scores for the three DL models are greater than 96%. Further, the OCC and CCVR scores achieved for the three DL models under consideration substantiate that they mostly focus on the choroid layer in making the decision. In particular, of the three DL models, EfficientNet-B3 is in close agreement with the clinician’s inference. The proposed DL-based framework demonstrated high detection accuracy as well as attention on the choroid layer, where EfficientNet-B3 reported superior performance. Our work assumes significance in bench-marking the automated choroid biomarker detection tools and facilitating high-throughput screening. Further, the methods proposed in this work can be adopted for evaluating the attention of DL-based approaches developed for other region-specific quality assessment tasks.

show abstract

Improved Fundus Image Quality Assessment: Augmenting Traditional Features with Structure Preserving ScatNet Features in Multicolor Space

Manne

Bashar

Vupparaboina

et al. 2021

2020 IEEE-EMBS Conference on Biomedical Engineering and Sciences (IECBES)

View full text Add to dashboard Cite

Diagnostic Quality Assessment of Ocular Fundus Photographs: Efficacy of Structure-Preserving ScatNet Features

Cited by 3 publications

References 9 publications

Deep Learning Based Diagnostic Quality Assessment of Choroidal OCT Features with Expert-Evaluated Explainability

Deep Learning Based Diagnostic Quality Assessment of Choroidal OCT Features with Expert-Evaluated Explainability

Deep learning based diagnostic quality assessment of choroidal OCT features with expert-evaluated explainability

Improved Fundus Image Quality Assessment: Augmenting Traditional Features with Structure Preserving ScatNet Features in Multicolor Space

Contact Info

Product

Resources

About