Noise-imitation learning: unpaired speckle noise reduction for optical coherence tomography.

Objective.Optical coherence tomography (OCT) is widely used in clinical practice for its non-invasive, high-resolution imaging capabilities. However, speckle noise inherent to its low coherence principle can degrade image quality and compromise diagnostic accuracy. While deep learning methods have shown promise in reducing speckle noise, obtaining well-registered image pairs remains challenging, leading to the development of unpaired methods. Despite their potential, existing unpaired methods suffer from redundancy in network structures or interaction mechanisms. Therefore, a more streamlined method for unpaired OCT denoising is essential.Approach.In this work, we propose a novel unpaired method for OCT image denoising, referred to as noise-imitation learning (NIL). NIL comprises three primary modules: the noise extraction module, which extracts noise features by denoising noisy images; the noise imitation module, which synthesizes noisy images and generates fake clean images; and the adversarial learning module, which differentiates between real and fake clean images through adversarial training. The complexity of NIL is significantly lower than that of previous unpaired methods, utilizing only one generator and one discriminator for training.Main results.By efficiently fusing unpaired images and employing adversarial training, NIL can extract more speckle noise information to enhance denoising performance. Building on NIL, we propose an OCT image denoising pipeline, NIL-NAFNet. This pipeline achieved PSNR, SSIM, and RMSE values of 31.27 dB, 0.865, and 7.00, respectively, on the PKU37 dataset. Extensive experiments suggest that our method outperforms state-of-the-art unpaired methods both qualitatively and quantitatively.Significance.These findings indicate that the proposed NIL is a simple yet effective method for unpaired OCT speckle noise reduction. The OCT denoising pipeline based on NIL demonstrates exceptional performance and efficiency. By addressing speckle noise without requiring well-registered image pairs, this method can enhance image quality and diagnostic accuracy in clinical practice.

Deep-KEDI: Deep learning-based zigzag generative adversarial network for encryption and decryption of medical images.

BACKGROUND: Medical imaging techniques have improved to the point where security has become a basic requirement for all applications to ensure data security and data transmission over the internet. However, clinical images hold personal and sensitive data related to the patients and their disclosure has a negative impact on their right to privacy as well as legal ramifications for hospitals. OBJECTIVE: In this research, a novel deep learning-based key generation network (Deep-KEDI) is designed to produce the secure key used for decrypting and encrypting medical images. METHODS: Initially, medical images are pre-processed by adding the speckle noise using discrete ripplet transform before encryption and are removed after decryption for more security. In the Deep-KEDI model, the zigzag generative adversarial network (ZZ-GAN) is used as the learning network to generate the secret key. RESULTS: The proposed ZZ-GAN is used for secure encryption by generating three different zigzag patterns (vertical, horizontal, diagonal) of encrypted images with its key. The zigzag cipher uses an XOR operation in both encryption and decryption using the proposed ZZ-GAN. Encrypting the original image requires a secret key generated during encryption. After identification, the encrypted image is decrypted using the generated key to reverse the encryption process. Finally, speckle noise is removed from the encrypted image in order to reconstruct the original image. CONCLUSION: According to the experiments, the Deep-KEDI model generates secret keys with an information entropy of 7.45 that is particularly suitable for securing medical images.

Ultrasound image denoising autoencoder model based on lightweight attention mechanism.

Background: The presence of noise in medical ultrasound images significantly degrades image quality and affects the accuracy of disease diagnosis. The convolutional neural network-denoising autoencoder (CNN-DAE) model extracts feature information by stacking regularly sized kernels. This results in the loss of texture detail, the over-smoothing of the image, and a lack of generalizability for speckle noise. Methods: A lightweight attention denoise-convolutional neural network (LAD-CNN) is proposed in the present study. Two different lightweight attention blocks (i.e., the lightweight channel attention (LCA) block and the lightweight large-kernel attention (LLA) block are concatenated into the downsampling stage and the upsampling stage, respectively. A skip connection is included before the upsampling layer to alleviate the problem of gradient vanishing during backpropagation. The effectiveness of our model was evaluated using both subjective visual effects and objective evaluation metrics. Results: With the highest peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) values at all noise levels, the proposed model outperformed the other models. In the test of brachial plexus ultrasound images, the average PSNR of our model was 0.15 higher at low noise levels and 0.33 higher at high noise levels than the suboptimal model. In the test of fetal ultrasound images, the average PSNR of our model was 0.23 higher at low noise levels and 0.20 higher at high noise levels than the suboptimal model. The statistical analysis showed that the p values were less than 0.05, which indicated a statistically significant difference between our model and the other models. Conclusions: The results of this study suggest that the proposed LAD-CNN model is more efficient in denoising and preserving image details than both conventional denoising algorithms and existing deep-learning algorithms.

Coupling speckle noise suppression with image classification for deep-learning-aided ultrasound diagnosis.

Objective. During deep-learning-aided (DL-aided) ultrasound (US) diagnosis, US image classification is a foundational task. Due to the existence of serious speckle noise in US images, the performance of DL models may be degraded. Pre-denoising US images before their use in DL models is usually a logical choice. However, our investigation suggests that pre-speckle-denoising is not consistently advantageous. Furthermore, due to the decoupling of speckle denoising from the subsequent DL classification, investing intensive time in parameter tuning is inevitable to attain the optimal denoising parameters for various datasets and DL models. Pre-denoising will also add extra complexity to the classification task and make it no longer end-to-end.Approach. In this work, we propose a multi-scale high-frequency-based feature augmentation (MSHFFA) module that couples feature augmentation and speckle noise suppression with specific DL models, preserving an end-to-end fashion. In MSHFFA, the input US image is first decomposed to multi-scale low-frequency and high-frequency components (LFC and HFC) with discrete wavelet transform. Then, multi-scale augmentation maps are obtained by computing the correlation between LFC and HFC. Last, the original DL model features are augmented with multi-scale augmentation maps.Main results. On two public US datasets, all six renowned DL models exhibited enhanced F1-scores compared with their original versions (by 1.31%-8.17% on the POCUS dataset and 0.46%-3.89% on the BLU dataset) after using the MSHFFA module, with only approximately 1% increase in model parameter count.Significance. The proposed MSHFFA has broad applicability and commendable efficiency and thus can be used to enhance the performance of DL-aided US diagnosis. The codes are available athttps://github.com/ResonWang/MSHFFA.

AdaRes: A deep learning-based model for ultrasound image denoising: Results of image quality metrics, radiomics, artificial intelligence, and clinical studies.

PURPOSE: The quality of ultrasound images is degraded by speckle and Gaussian noises. This study aims to develop a deep-learning (DL)-based filter for ultrasound image denoising. METHODS: A novel DL-based filter using adaptive residual (AdaRes) learning was proposed. Five image quality metrics (IQMs) and 27 radiomics features were used to evaluate denoising results. The effect of our proposed filter, AdaRes, on four pre-trained convolutional neural network (CNN) classification models and three radiologists was assessed. RESULTS: AdaRes filter was tested on both natural and ultrasound image databases. IQMs results indicate that AdaRes could remove noises in three different noise levels with the highest performances. In addition, a radiomics study proved that AdaRes did not distort tissue textures and it could preserve most radiomics features. AdaRes could also improve the performance classification using CNNs in different settings. Finally, AdaRes also improved the mean overall performance (AUC) of three radiologists from 0.494 to 0.702 in the classification of benign and malignant lesions. CONCLUSIONS: AdaRes filtered out noises on ultrasound images more effectively and can be used as an auxiliary preprocessing step in computer-aided diagnosis systems. Radiologists may use it to remove unwanted noises and improve the ultrasound image quality before the interpretation.

Make It Less Complex: Autoencoder for Speckle Noise Removal-Application to Breast and Lung Ultrasound.

Ultrasound (US) imaging is used in the diagnosis and monitoring of COVID-19 and breast cancer. The presence of Speckle Noise (SN) is a downside to its usage since it decreases lesion conspicuity. Filters can be used to remove SN, but they involve time-consuming computation and parameter tuning. Several researchers have been developing complex Deep Learning (DL) models (150,000-500,000 parameters) for the removal of simulated added SN, without focusing on the real-world application of removing naturally occurring SN from original US images. Here, a simpler (<30,000 parameters) Convolutional Neural Network Autoencoder (CNN-AE) to remove SN from US images of the breast and lung is proposed. In order to do so, simulated SN was added to such US images, considering four different noise levels (σ = 0.05, 0.1, 0.2, 0.5). The original US images (N = 1227, breast + lung) were given as targets, while the noised US images served as the input. The Structural Similarity Index Measure (SSIM) and Peak Signal-to-Noise Ratio (PSNR) were used to compare the output of the CNN-AE and of the Median and Lee filters with the original US images. The CNN-AE outperformed the use of these classic filters for every noise level. To see how well the model removed naturally occurring SN from the original US images and to test its real-world applicability, a CNN model that differentiates malignant from benign breast lesions was developed. Several inputs were used to train the model (original, CNN-AE denoised, filter denoised, and noised US images). The use of the original US images resulted in the highest Matthews Correlation Coefficient (MCC) and accuracy values, while for sensitivity and negative predicted values, the CNN-AE-denoised US images (for higher σ values) achieved the best results. Our results demonstrate that the application of a simpler DL model for SN removal results in fewer misclassifications of malignant breast lesions in comparison to the use of original US images and the application of the Median filter. This shows that the use of a less-complex model and the focus on clinical practice applicability are relevant and should be considered in future studies.

Toward cell nuclei precision between OCT and H&E images translation using signal-to-noise ratio cycle-consistency.

Medical image-to-image translation is often difficult and of limited effectiveness due to the differences in image acquisition mechanisms and the diverse structure of biological tissues. This work presents an unpaired image translation model between in-vivo optical coherence tomography (OCT) and ex-vivo Hematoxylin and eosin (H&E) stained images without the need for image stacking, registration, post-processing, and annotation. The model can generate high-quality and highly accurate virtual medical images, and is robust and bidirectional. Our framework introduces random noise to (1) blur redundant features, (2) defend against self-adversarial attacks, (3) stabilize inverse conversion, and (4) mitigate the impact of OCT speckles. We also demonstrate that our model can be pre-trained and then fine-tuned using images from different OCT systems in just a few epochs. Qualitative and quantitative comparisons with traditional image-to-image translation models show the robustness of our proposed signal-to-noise ratio (SNR) cycle-consistency method.

Medical Image Despeckling Using the Invertible Sparse Fuzzy Wavelet Transform with Nature-Inspired Minibatch Water Wave Swarm Optimization.

Speckle noise is a pervasive problem in medical imaging, and conventional methods for despeckling often lead to loss of edge information due to smoothing. To address this issue, we propose a novel approach that combines a nature-inspired minibatch water wave swarm optimization (NIMWVSO) framework with an invertible sparse fuzzy wavelet transform (ISFWT) in the frequency domain. The ISFWT learns a non-linear redundant transform with a perfect reconstruction property that effectively removes noise while preserving structural and edge information in medical images. The resulting threshold is then used by the NIMWVSO to further reduce multiplicative speckle noise. Our approach was evaluated using the MSTAR dataset, and objective functions were based on two contrasting reference metrics, namely the peak signal-to-noise ratio (PSNR) and the mean structural similarity index metric (MSSIM). Our results show that the suggested approach outperforms modern filters and has significant generalization ability to unknown noise levels, while also being highly interpretable. By providing a new framework for despeckling medical images, our work has the potential to improve the accuracy and reliability of medical imaging diagnosis and treatment planning.

Optical coherence tomography image despeckling based on tensor singular value decomposition and fractional edge detection.

Optical coherence tomography (OCT) imaging is a technique that is frequently used to diagnose medical conditions. However, coherent noise, sometimes referred to as speckle noise, can dramatically reduce the quality of OCT images, which has an adverse effect on how OCT images are used. In order to enhance the quality of OCT images, a speckle noise reduction technique is developed, and this method is modelled as a low-rank tensor approximation issue. The grouped 3D tensors are first transformed into the transform domain using tensor singular value decomposition (t-SVD). Then, to cut down on speckle noise, transform coefficients are thresholded. Finally, the inverse transform can be used to produce images with speckle suppression. To further enhance the despeckling results, a feature-guided thresholding approach based on fractional edge detection and an adaptive backward projection technique are also presented. Experimental results indicate that the presented algorithm outperforms several comparison methods in relation to speckle suppression, objective metrics, and edge preservation.

Image Noise Removal in Ultrasound Breast Images Based on Hybrid Deep Learning Technique.

Rapid improvements in ultrasound imaging technology have made it much more useful for screening and diagnosing breast problems. Local-speckle-noise destruction in ultrasound breast images may impair image quality and impact observation and diagnosis. It is crucial to remove localized noise from images. In the article, we have used the hybrid deep learning technique to remove local speckle noise from breast ultrasound images. The contrast of ultrasound breast images was first improved using logarithmic and exponential transforms, and then guided filter algorithms were used to enhance the details of the glandular ultrasound breast images. In order to finish the pre-processing of ultrasound breast images and enhance image clarity, spatial high-pass filtering algorithms were used to remove the extreme sharpening. In order to remove local speckle noise without sacrificing the image edges, edge-sensitive terms were eventually added to the Logical-Pool Recurrent Neural Network (LPRNN). The mean square error and false recognition rate both fell below 1.1% at the hundredth training iteration, showing that the LPRNN had been properly trained. Ultrasound images that have had local speckle noise destroyed had signal-to-noise ratios (SNRs) greater than 65 dB, peak SNR ratios larger than 70 dB, edge preservation index values greater than the experimental threshold of 0.48, and quick destruction times. The time required to destroy local speckle noise is low, edge information is preserved, and image features are brought into sharp focus.

A New Estimation Scheme for Improving the Performance of Shear Wave Elasticity Imaging.

Shear wave velocity (SWV) reconstruction based on time-of-flight (TOF) is widely adopted to realize shear wave elasticity imaging (SWEI). It typically breaks down the reconstruction of a SWV image into many kernels and treats them independently. We hypothesized that information exchange among kernels improves the performance of SWEI. Therefore, we propose the approach of iterative re-weighted least squares based on inter-kernel communication (IKC-IRLS). We also hypothesized that time-to-peak (TTP) is superior to cross-correlation (CC) in visualizing small targets because TTP uses higher shear wave frequencies than CC. To examine the hypotheses, IKC-IRLS was combined with TTP data and compared with four established methods. The five methods were tested by imaging several small-size stiff targets (2.5, 4.0 and 6.4 mm in diameter) using different kernel sizes in the simulation and real experiments. The results indicate that the IKC-IRLS approach can mitigate speckle noise and is robust to TTP outliers. Consequently, the proposed method achieves the highest contrast-to-noise ratio and the lowest mean absolute percentage error of target in almost all tested cases.

Non-Local Mean Denoising Algorithm Based on Fractional Compact Finite Difference Scheme Effectively Reduces Speckle Noise in Optical Coherence Tomography Images.

Optical coherence tomography (OCT) is used in various fields such, as medical diagnosis and material inspection, as a non-invasive and high-resolution optical imaging modality. However, an OCT image is damaged by speckle noise during its generation, thus reducing the image quality. To address this problem, a non-local means (NLM) algorithm based on the fractional compact finite difference scheme (FCFDS) is proposed to remove the speckle noise in OCT images. FCFDS uses more local pixel information when compared to integer-order difference operators. The FCFDS operator is introduced into the NLM algorithm to construct a high-precision weight calculation so that the proposed algorithm can effectively reduce the speckle noise in the OCT images. Experiments on simulations and real OCT images show that the proposed method is comparable to other state-of-the-art despeckling methods and can substantially reduce noise and preserve image details such as edges and structures. Speckle noise removal can further promote the application of the proposed algorithm in medical diagnosis and industrial detection, as it has key research value.

Deep Learning-Based 3D Measurements with Near-Infrared Fringe Projection.

Fringe projection profilometry (FPP) is widely applied to 3D measurements, owing to its advantages of high accuracy, non-contact, and full-field scanning. Compared with most FPP systems that project visible patterns, invisible fringe patterns in the spectra of near-infrared demonstrate fewer impacts on human eyes or on scenes where bright illumination may be avoided. However, the invisible patterns, which are generated by a near-infrared laser, are usually captured with severe speckle noise, resulting in 3D reconstructions of limited quality. To cope with this issue, we propose a deep learning-based framework that can remove the effect of the speckle noise and improve the precision of the 3D reconstruction. The framework consists of two deep neural networks where one learns to produce a clean fringe pattern and the other to obtain an accurate phase from the pattern. Compared with traditional denoising methods that depend on complex physical models, the proposed learning-based method is much faster. The experimental results show that the measurement accuracy can be increased effectively by the presented method.

Enhanced Child Care: Contrast Correction for Pediatric Hip Ultrasound Using Hyperanalytic Wavelets.

(1) Background: The prevention of critical situations is a key ability in medicine. Hip ultrasound for neonates is a standard procedure to prevent later critical outcomes, such as hip dysplasia. Additionally, the SARS-CoV-2 pandemic has put worldwide stress upon healthcare units, resulting often in a lack of sufficient medical personnel. This work aims to develop solutions to ease and speed up the process of coming to a correct diagnosis. (2) Methods: Traditional medical procedures are envisaged, but they are enhanced to reduce diagnosing errors due to the movements of the neonates. Echographic noise filtering and contrast correction methods are implemented the Hyperanalytic Wavelet Transform, combined with an adaptive Soft Thresholding Filter. The algorithm is tailored to infants' structure and is tested on real ultrasounds provided by the "Victor Babes" University of Medicine and Pharmacy. Denoising and contrast correction problems are targeted. (3) Results: In available clinical cases, the noise affecting the image was reduced and the contrast was enhanced. (4) Discussion: We noticed that a significant amount of noise can be added to the image, as the patients are neonates and can hardly avoid movements. (5) Conclusions: The algorithm is personalized with no fixed reference value. Any device easing the clinical procedures of physicians has a practical medical application.

DHNet: High-resolution and hierarchical network for cross-domain OCT speckle noise reduction.

PURPOSE: Optical coherence tomography (OCT) imaging uses the principle of Michelson interferometry to obtain high-resolution images by coherent superposing of multiple forward and backward scattered light waves with random phases. This process inevitably produces speckle noise that severely compromises visual quality of OCT images and degrades performances of subsequent image analysis tasks. In addition, datasets obtained by different OCT scanners have distribution shifts, making a speckle noise suppression model difficult to be generalized across multiple datasets. In order to solve the above issues, we propose a novel end-to-end denoising framework for OCT images collected by different scanners. METHODS: The proposed model utilizes the high-resolution network (HRNet) as backbone for image restoration, which reconstructs high-fidelity images by maintaining high-resolution representations throughout the entire learning process. To compensate distribution shifts among datasets collected by different scanners, we develop a hierarchical adversarial learning strategy for domain adaption. The proposed model is trained using datasets with clean ground truth produced by two commercial OCT scanners, and then applied to suppress speckle noise in OCT images collected by our recently developed OCT scanner, BV-1000 (China Bigvision Corporation). We name the proposed model as DHNet (Double-H-Net, High-resolution and Hierarchical Network). RESULTS: We compare DHNet with state-of-the-art methods and experiment results show that DHNet improves signal-to-noise ratio by a large margin of 18.137 dB as compared to the best of our previous method. In addition, DHNet achieves a testing time of 25 ms, which satisfies the real-time processing requirement for the BV-1000 scanner. We also conduct retinal layer segmentation experiment on OCT images before and after denoising and show that DHNet can also improve segmentation. CONCLUSIONS: The proposed DHNet can compensate domain shifts between different data sets while significantly improve speckle noise suppression. The HRNet backbone is utilized to carry low- and high-resolution information to recover fidelity images. Domain adaptation is achieved by a hierarchical module through adversarial learning. In addition, DHNet achieved a testing time of 25 ms, which satisfied the real-time processing requirement.

Improved Weighted Non-Local Mean Filtering Algorithm for Laser Image Speckle Suppression.

Laser speckle noise caused by coherence between lasers greatly influences the produced image. In order to suppress the effect of laser speckles on images, in this paper we set up a combination of a laser-structured light module and an infrared camera to acquire laser images, and propose an improved weighted non-local mean (IW-NLM) filtering method that adopts an SSI-based adaptive h-solving method to select the optimal h in the weight function. The analysis shows that the algorithm not only denoises the laser image but also smooths pixel jumps in the image, while preserving the image details. The experimental results show that compared with the original laser image, the equivalent number of looks (ENL) index of the IW-NLM filtered image improved by 0.80%. The speckle suppression index (SSI) of local images dropped from 4.69 to 2.55%. Compared with non-local mean filtering algorithms, the algorithm proposed in this paper is an improvement and provides more accurate data support for subsequent image processing analysis.

Bayesian Framework-Based Adaptive Hybrid Filtering for Speckle Noise Reduction in Ultrasound Images Via Lion Plus FireFly Algorithm.

The existence of speckle noise in ultrasound (US) image processing distorts the image quality and also hinders the development of systematic approaches for US images. Numerous de-speckling schemes were established to date that concern speckle reduction; however, the models suffer from demerits like computational time, computational complexity, etc., that are to be rectified as soon as possible. This compulsion takes to the introduction of a new de-speckling model via an adaptive hybrid filter model that includes four filters like guided filter (GF), speckle-reducing bilateral filter (SRBF), rotation invariant bilateral nonlocal means filter (RIBNLM), and median filter (MF) respectively. Moreover, the novelty goes under the selection of optimal filter coefficients that make the process effective. Bayesian-based neural network is used to predict the appropriate filter coefficients, where the training library is constructed with the optimal coefficients. Along with this, the selection of optimal filter coefficients is done under the defined objective function using a new hybrid algorithm termed as Randomized FireFly (FF) update in Lion Algorithm (RFU-LA) that hybrids the concept of both LA and FF, respectively. Finally, the performance of the proposed de-speckling model is compared over that of other conventional models with respect to different performance measures. Accordingly, from the analysis, the mean MAPE of the proposed method are 39.13% and 49.28% higher than those of the wavelet filtering and hybrid filtering schemes for a noise variance of 0.1.

Adversarial Attack and Defence through Adversarial Training and Feature Fusion for Diabetic Retinopathy Recognition.

Due to the rapid growth in artificial intelligence (AI) and deep learning (DL) approaches, the security and robustness of the deployed algorithms need to be guaranteed. The security susceptibility of the DL algorithms to adversarial examples has been widely acknowledged. The artificially created examples will lead to different instances negatively identified by the DL models that are humanly considered benign. Practical application in actual physical scenarios with adversarial threats shows their features. Thus, adversarial attacks and defense, including machine learning and its reliability, have drawn growing interest and, in recent years, has been a hot topic of research. We introduce a framework that provides a defensive model against the adversarial speckle-noise attack, the adversarial training, and a feature fusion strategy, which preserves the classification with correct labelling. We evaluate and analyze the adversarial attacks and defenses on the retinal fundus images for the Diabetic Retinopathy recognition problem, which is considered a state-of-the-art endeavor. Results obtained on the retinal fundus images, which are prone to adversarial attacks, are 99% accurate and prove that the proposed defensive model is robust.

Robust three-dimensional registration on optical coherence tomography angiography for speckle reduction and visualization.

BACKGROUND: In the clinical applications of optical coherence tomography angiography (OCTA), the repeated scanning and averaging method can provide better contrast with reduced speckle noises in the final results, which are useful for visualizing and quantifying vascular components with high accuracy, reproducibility, and reliability. However, the inevitable patient motion presents a challenge to this method. The objective of this study is to meet this challenge by introducing a 3D registration method to register optical coherence tomography (OCT)/OCTA scans for precise volume averaging of multiple scans to improve the signal-to-noise ratio (SNR) and increase quantification accuracy. METHODS: The proposed method utilized both rigid affine transformation and non-rigid B-spline transformation in which their parameters were optimized and calculated by the average stochastic gradient descent on OCT structural images. In addition, we also introduced a multi-level resolution approach to further improve the robustness and computational speed of our proposed method. The imaging performance was tested on in vivo imaging of human skin and eye and assessed by SNR, peak signal-to-noise ratio (PSNR) and normalized correlation coefficient (NCC). RESULTS: Five subjects were enrolled in this study for obtaining in vivo images of skin and retina. The proposed registration and averaging method provided substantial improvements of the imaging performance in terms of vessel connectivity and signal to noise ratio. The increase of repeated volume numbers in the averaging improves all the metrics assessed, i.e., SNR, PSNR and NCC. An improvement of the SNR from 10 to 40 dB after 10 repeated volumetric averaging was achieved. CONCLUSIONS: The proposed 3D registration and averaging method is effective in reducing speckle noises and suppressing motion artifacts, thereby improving SNR, PSNR and NCC metrics for final averaged images. It is expected that the proposed algorithm would be practically useful in better visualization and more reliable quantification of in vivo OCT and OCTA data, which would be beneficial to OCT clinical applications.

Towards Automated Eye Diagnosis: An Improved Retinal Vessel Segmentation Framework Using Ensemble Block Matching 3D Filter.

Automated detection of vision threatening eye disease based on high resolution retinal fundus images requires accurate segmentation of the blood vessels. In this regard, detection and segmentation of finer vessels, which are obscured by a considerable degree of noise and poor illumination, is particularly challenging. These noises include (systematic) additive noise and multiplicative (speckle) noise, which arise due to various practical limitations of the fundus imaging systems. To address this inherent issue, we present an efficient unsupervised vessel segmentation strategy as a step towards accurate classification of eye diseases from the noisy fundus images. To that end, an ensemble block matching 3D (BM3D) speckle filter is proposed for removal of unwanted noise leading to improved detection. The BM3D-speckle filter, despite its ability to recover finer details (i.e., vessels in fundus images), yields a pattern of checkerboard artifacts in the aftermath of multiplicative (speckle) noise removal. These artifacts are generally ignored in the case of satellite images; however, in the case of fundus images, these artifacts have a degenerating effect on the segmentation or detection of fine vessels. To counter that, an ensemble of BM3D-speckle filter is proposed to suppress these artifacts while further sharpening the recovered vessels. This is subsequently used to devise an improved unsupervised segmentation strategy that can detect fine vessels even in the presence of dominant noise and yields an overall much improved accuracy. Testing was carried out on three publicly available databases namely Structured Analysis of the Retina (STARE), Digital Retinal Images for Vessel Extraction (DRIVE) and CHASE_DB1. We have achieved a sensitivity of 82.88, 81.41 and 82.03 on DRIVE, SATARE, and CHASE_DB1, respectively. The accuracy is also boosted to 95.41, 95.70 and 95.61 on DRIVE, SATARE, and CHASE_DB1, respectively. The performance of the proposed methods on images with pathologies was observed to be more convincing than the performance of similar state-of-the-art methods.