Microaneurysms Research Articles

An efficient method for early stage detection of diabetic retinopathy

Diabetic Retinopathy (DR) is one of the leading causes of blindness. The early detection and treatment of DR is significant to save the human vision. The presence of microaneurysms (MAs) is the first sign of the disease. The correct identification of MAs is an essential for finding of DR at the early stages. In this paper, we propose a three phase system for efficient recognition of MAs. The tentative MA lesions are recovered from the fundus image in the first stage. To accurately classify an extracted candidate region into MA or non-MA, the second stage prepares an attribute vector for each tentative MA lesion based on shape, intensity and statistical properties. The third stage is a classification step to classify as MAs and Non-MAs for early stage detection of DR. We present a holoentropy enabled decision tree classifier which combines entropy and total correlation. The best feature for decision tree is selected based on holoentropy to enhance the correctness of the classification. The implemented system is experimented using fundus image database DIARETDB1. The proposed method achieved an overall accuracy of 97.67%.The proposed system has detected the MAs with higher performance using simple features and holoentropy based decision tree classifier. The proposed system is suitable for early stage detection of DR through the accurate identification of MAs.

EFFICIENT DETECTION OF MICRO-ANEURYSMS IN RETINAL IMAGES FOR EARLY DIAGNOSIS OF DIABETIC RETINOPATHY

Diabetic retinopathy (DR) is a serious complication of diabetes mellitus and one of the major causes of blindness worldwide. As the number of diabetic patients increases, early detection of DR for regular screening can prevent loss of vision and blindness. The development of algorithms for detecting dark lesions may turn very useful in the early diagnosis and screening of retinopathy diseases. In this paper, a novel hybrid algorithm for the microaneurysms (MAs) detection is developed. This task is based on mathematical morphology, which is followed by a classification step. Although some algorithms have been developed, the accurate detection of MAs in color retinal images is still a challenging problem. The proposed approach aims to increase the number of true positives and minimize the false positives compared with methods developed in the literature. The new approach is tested on a set of 219 ophthalmologic images. A total of 12 microaneurysm features are considered in this study and selected for KNN classification. The validity of detection process is checked through comparisons at the pixel level with ophthalmologists’ hand-drawn ground truth. Sensitivity, specificity, prediction rate, and accuracy achieved by the proposed approach are 98.13%, 99.71%, 99.63% and 99.01% respectively.

A novel microaneurysms detection approach based on convolutional neural networks with reinforcement sample learning algorithm.

Microaneurysms (MAs) are known as early signs of diabetic-retinopathy which are called red lesions in color fundus images. Detection of MAs in fundus images needs highly skilled physicians or eye angiography. Eye angiography is an invasive and expensive procedure. Therefore, an automatic detection system to identify the MAs locations in fundus images is in demand. In this paper, we proposed a system to detect the MAs in colored fundus images. The proposed method composed of three stages. In the first stage, a series of pre-processing steps are used to make the input images more convenient for MAs detection. To this end, green channel decomposition, Gaussian filtering, median filtering, back ground determination, and subtraction operations are applied to input colored fundus images. After pre-processing, a candidate MAs extraction procedure is applied to detect potential regions. A five-stepped procedure is adopted to get the potential MA locations. Finally, deep convolutional neural network (DCNN) with reinforcement sample learning strategy is used to train the proposed system. The DCNN is trained with color image patches which are collected from ground-truth MA locations and non-MA locations. We conducted extensive experiments on ROC dataset to evaluate of our proposal. The results are encouraging.

An ensemble deep learning based approach for red lesion detection in fundus images

Background and objectives: Diabetic retinopathy (DR) is one of the leading causes of preventable blindness in the world. Its earliest sign are red lesions, a general term that groups both microaneurysms (MAs) and hemorrhages (HEs). In daily clinical practice, these lesions are manually detected by physicians using fundus photographs. However, this task is tedious and time consuming, and requires an intensive effort due to the small size of the lesions and their lack of contrast. Computer-assisted diagnosis of DR based on red lesion detection is being actively explored due to its improvement effects both in clinicians consistency and accuracy. Moreover, it provides comprehensive feedback that is easy to assess by the physicians. Several methods for detecting red lesions have been proposed in the literature, most of them based on characterizing lesion candidates using hand crafted features, and classifying them into true or false positive detections. Deep learning based approaches, by contrast, are scarce in this domain due to the high expense of annotating the lesions manually.Methods: In this paper we propose a novel method for red lesion detection based on combining both deep learned and domain knowledge. Features learned by a convolutional neural network (CNN) are augmented by incorporating hand crafted features. Such ensemble vector of descriptors is used afterwards to identify true lesion candidates using a Random Forest classifier.Results: We empirically observed that combining both sources of information significantly improve results with respect to using each approach separately. Furthermore, our method reported the highest performance on a per-lesion basis on DIARETDB1 and e-ophtha, and for screening and need for referral on MESSIDOR compared to a second human expert.Conclusions: Results highlight the fact that integrating manually engineered approaches with deep learned features is relevant to improve results when the networks are trained from lesion-level annotated data. An open source implementation of our system is publicly available at https://github.com/ignaciorlando/red-lesion-detection.

Statistical Geometrical Features for Microaneurysm Detection

Automated microaneurysm (MA) detection is still an open challenge due to its small size and similarity with blood vessels. In this paper, we present a novel method which is simple, efficient, and real-time for segmenting and detecting MA in color fundus images (CFI). To do this, a novel set of features based on statistics of geometrical properties of connected regions, that can easily discriminate lesion and non-lesion pixels are used. For large-scale evaluation proposed method is validated on DIARETDB1, ROC, STARE, and MESSIDOR dataset. It proves robust with respect to different image characteristics and camera settings. The best performance was achieved on per-image evaluation on DIARETDB1 dataset with sensitivity of 88.09 at 92.65% specificity which is quite encouraging for clinical use.

Non-rigid registration of Fluorescein Angiography and Optical Coherence Tomography via scanning laser ophthalmoscope imaging.

Fluorescein Angiography (FA) imaging is the gold standard technique for neurovascular imaging regarding assessing neurovascular diseases such as Diabetic Retinopathy (DR) and Diabetic Macular Edema (DME). On the other hand, as FA imaging is invasive and does not provide any depth information, Optical Coherence Tomography (OCT) imaging technique is a good complementary for it in diagnosis process. To correlate the information of both FA and OCT images, an image alignment/registration process is needed. In absence of an automatic registration software, the clinician should do intuitive comparison to integrate these data which is a subjective and time consuming process. In this paper, we demonstrate a non-rigid registration method called multi-step correlation-based registration algorithm to automatically register FA and OCT images together. Our algorithm consists of two steps including rigid/global and non-rigid/local registration. We evaluate our algorithm's performance by labeling Micro-Aneurysm (MA) spots -hallmarks of DR- on FA images and determining MA regions on OCT B-scans after registration. Our Results show that our algorithm performs accurately regarding registration of FA images and OCT B-scans.

Diabetic Microaneurysms Internal Reflectivity on Spectral-Domain Optical Coherence Tomography and Optical Coherence Tomography Angiography Detection

To correlate the appearance of microaneurysms (MAs) on structural spectral-domain optical coherence tomography (SDOCT) with their detection on OCT angiography (OCTA) in patients with nonproliferative diabetic retinopathy (NPDR). Interinstrument reliability study. Sixteen patients with NPDR without macular edema underwent SDOCT and OCTA. To compare MAs seen on OCTA with those on SDOCT, we superimposed the OCTA superficial capillary plexus (SCP) vascular landmarks onto those of the near infrared. Two observers masked to patient groupings evaluated reflectivity of MAs on SDOCT scans, graded as hyporeflective, moderate, or hyperreflective, and their visualization at the level of SCP and deep capillary plexus (DCP) on OCTA. Among 145 MAs imaged with SDOCT, 47 (32.4%) appeared as hyperreflective, 71 (49.0%) as moderately reflective, and 27 (18.6%) as hyporeflective. After excluding 3 eyes (10 MAs) because of poor-quality OCTA scans, 135 MAs were evaluated on OCTA; 76 (56.3%) were visible only in the DCP, 9 (6.7%) only in the SCP, 29 (21.5%) were visible in both SCP and DCP; and 21 (15.6%) were not visible on OCTA. Compared with MAs with hyperreflectivity or moderate reflectivity, MAs with hyporeflectivity on structural SDOCT were significantly less likely to be detected on OCTA (odds ratio [OR]: 4.6; 95% confidence interval [CI]: 1.5-14.0, P= .008; and OR: 4.2, 95% CI: 1.2-14.2, P= .022, respectively). MAs that appear hyporeflective on structural SDOCT have a lower detection rate on OCTA. The results of this study may help further understand the different blood flow dynamics pattern in MAs.

Automated quantitative characterisation of retinal vascular leakage and microaneurysms in ultra-widefield fluorescein angiography

Ultra-widefield fluorescein angiography (UWFA) is an emerging imaging modality used to characterise pathology in the retinal vasculature such as microaneurysms (MAs) and vascular leakage. Despite its potential value for diagnosis...

A Quantitative Approach to Predict Differential Effects of Anti-VEGF Treatment on Diffuse and Focal Leakage in Patients with Diabetic Macular Edema: A Pilot Study

PurposeWe use semiautomated segmentation of fluorescein angiography (FA) to determine whether anti-vascular endothelial growth factor (VEGF) treatment for diabetic macular edema (DME) differentially affects microaneurysm (MA)–associated leakage, termed focal leakage, versus non-MA–associated leakage, termed diffuse leakage.MethodsWe performed a retrospective study of 29 subjects treated with at least three consecutive injections of anti-VEGF agents for DME (mean 4.6 injections; range, 3–10) who underwent Heidelberg FA before and after anti-VEGF therapy. Inclusion criteria were macula center involving DME and at least 3 consecutive anti-VEGF injections. Exclusion criteria were macular edema due to cause besides DME, anti-VEGF within 3 months of initial FA, concurrent treatment for DME besides anti-VEGF, and macular photocoagulation within 1 year. At each time point, total leakage was semiautomatically segmented using a modified version of our previously published software. Microaneurysms were identified by an expert grader and leakage within a 117 μm radius of each MA was classified as focal leakage. Remaining leakage was classified as diffuse leakage. The absolute and percent changes in total, diffuse, and focal leakage were calculated for each subject.ResultsMean pretreatment total leakage was 8.2 mm2 and decreased by a mean of 40.1% (P < 0.0001; 95% confidence interval [CI], [−28.6, −52.5]) following treatment. Diffuse leakage decreased by a mean of 45.5% (P < 0.0001; 95% CI, [−31.3, −59.6]) while focal leakage decreased by 17.9% (P = 0.02; 95% CI, [−1.0, −34.8]). The difference in treatment response between focal and diffuse leakage was statistically significant (P = 0.01).ConclusionsAnti-VEGF treatment for DME results in decreased diffuse leakage but had relatively little effect on focal leakage as assessed by FA. This suggests that diffuse leakage may be a marker of VEGF-mediated pathobiology. Patients with predominantly focal leakage may be less responsive to anti-VEGF therapy.Translational RelevanceFluorescein angiography can define focal and diffuse subtypes of diabetic macular edema and these may respond differently to anti-VEGF treatment.

Clinical features and surgical treatment of hemorrhagic type moyamoya disease

Objective To investigate the clinical characteristics of hemorrhagic type moyamoya disease and its outcomes after revascularization. Methods The data of 46 patients with hemorrhagic moyamoya disease, including clinical features and surgical outcomes, were retrospectively reviewed. Among them, 42 cases were treated at Beijing Tiantan Hospital from August 2007 to November 2011. The other 4 cases were managed at the People's Liberation Army (PLA) 174 Hospital from August 2011 to June 2014. In this series of 46 patients, the operation of revascularization was performed in 50 hemispheres. Among them, direct superficial temporal artery to middle cerebral artery (STA-MCA)bypass surgery was conducted in 28 hemispheres, the encephalo-duro-arterio-synangiosis (EDAS) was performed in 10 hemispheres, and 12 hemispheres underwent a combination of STA-MCA bypass surgery and EDAS. Results The mean follow-up period was 52.5±6.5 months with a range of 18 ~ 87 months. Among the 19 cases with preoperative neurological deficiency, the symptoms were improved in 11cases, remained unchanged in 6 cases and became worse in 2 cases due to rebleeding. Among the preoperatively asymptomatic 27 cases, good outcomes were observed in 24 cases and rebleeding occurred in the non-surgical hemispheres of the remaining 3 cases. Follow-up CT Perfusion (CTP) and/or SPECT were performed in 25 cases and demonstrated significant improvement of cerebral blood perfusion in 20 cases and no remarkable change in 5 cases. Follow-up (at least 3 months post operation) DSA in the 15 patients who underwent STA-MCA bypass showed complete patency of the anastomosis and diminishing of the moyamoya vessels. In addition, 2 concurrent microaneurysms was found to have disappeared after the direct bypass surgery. Rebleeding developed in 2 (4%) hemispheres (post EDAS) in the surgical group and 3 (7%) hemispheres in the nonsurgical group. There was no statistical difference in the rate of rebleeding rate between the 2 groups(P>0.05). Conclusion The surgery of revascularization could improve the symptoms and cerebral perfusion for the patients with hemorrhagic moyamoya disease. Key words: Moyamoya disease; Hemorrhage; Cerebral revascularization; Treatment outcome

Detection of microaneurysms in retinal images using an ensemble classifier

Abstract This paper introduces, and reports on the performance of, a novel combination of algorithms for automated microaneurysm (MA) detection in retinal images. The presence of MAs in retinal images is a pathognomonic sign of Diabetic Retinopathy (DR) which is one of the leading causes of blindness amongst the working age population. An extensive survey of the literature is presented and current techniques in the field are summarised. The proposed technique first detects an initial set of candidates using a Gaussian Matched Filter and then classifies this set to reduce the number of false positives. A Tree Ensemble classifier is used with a set of 70 features (the most commons features in the literature). A new set of 32 MA groundtruth images (with a total of 256 labelled MAs) based on images from the MESSIDOR dataset is introduced as a public dataset for benchmarking MA detection algorithms. We evaluate our algorithm on this dataset as well as another public dataset (DIARETDB1 v2.1) and compare it against the best available alternative. Results show that the proposed classifier is superior in terms of eliminating false positive MA detection from the initial set of candidates. The proposed method achieves an ROC score of 0.415 compared to 0.2636 achieved by the best available technique. Furthermore, results show that the classifier model maintains consistent performance across datasets, illustrating the generalisability of the classifier and that overfitting does not occur.

Automatic Microaneurysm Detection Using the Sparse Principal Component Analysis-Based Unsupervised Classification Method

Since microaneurysms (MAs) can be seen as the earliest lesions in diabetic retinopathy, its detection plays a critical role in the diabetic retinopathy diagnosis. In recent years, many machine-learning methods have been developed for MA detection. Generally, MA candidates are first identified and then a set of features for these candidates are extracted. Finally, machine-learning methods are applied for candidate classification. In this paper, we present a novel unsupervised classification method based on sparse posterior cerebral artery (PCA) for MA detection. Since it does not have to consider a non-MA training set, the class imbalance problem can be avoided. Furthermore, effective features can be selected due to the characteristic of sparse PCA, which combines the elastic net penalty with the PCA. Meanwhile, a single $T^{2}$ statistic is introduced, and the control limit can be determined for distinguishing true MAs from spurious candidates automatically. Experiment results on the retinopathy online challenge competition database show the effectiveness of our proposed method.

Automatic Microaneurysms Detection Based on Multifeature Fusion Dictionary Learning.

Recently, microaneurysm (MA) detection has attracted a lot of attention in the medical image processing community. Since MAs can be seen as the earliest lesions in diabetic retinopathy, their detection plays a critical role in diabetic retinopathy diagnosis. In this paper, we propose a novel MA detection approach named multifeature fusion dictionary learning (MFFDL). The proposed method consists of four steps: preprocessing, candidate extraction, multifeature dictionary learning, and classification. The novelty of our proposed approach lies in incorporating the semantic relationships among multifeatures and dictionary learning into a unified framework for automatic detection of MAs. We evaluate the proposed algorithm by comparing it with the state-of-the-art approaches and the experimental results validate the effectiveness of our algorithm.

1892 and the tribulations of Joseph Babinski

In the early 1880s, Joseph Babinski was appointed as Chef de Clinique under Jean-Martin Charcot at Salpétrière, Paris, in France. He appeared for the post of Professeur Agrégé, the pinnacle of academic distinction in France in 1892. Charles Bouchard, the earliest pupil of Charcot, who described the Charcot-Bouchard aneurysm along with his master, was the Chief of the Board of Jury. Charcot and Bouchard did not see eye to eye in the later period, and when German Sée, an external examiner did not join the board of examiners following illness, Bouchard, instead of seeking the assistance of an alternate examiner, employed the special right of Vote of Absence. Babinski and all other pupils of Charcot were unsuccessful, while those of Bouchard, came out with flying colors. An embittered Babinski, along with some of the other unsuccessful candidates, appeared before the Ministry of Public Assistance after a protracted legal battle for nearly 2 years but lost the case. They were even ordered by the court to pay on behalf of the the successful candidates for the legal battle. Babinski never sat for the examination again and this is the story of one of the most ignominious episodes in an examination anywhere.

A Novel Approach for Detection of Microaneurysms in Diabetic Retinopathy Disease from Retinal Fundus Images

Diabetic Retinopathy (DR) is a leading cause of blindness in human beings aged between 20 to 74 years. It has a great influence on the patient and society because it normally influences humans in their most gainful years. Early detection in DR is very difficult which is not detected by human beings. Many algorithms and techniques were established to detect DR. These techniques faced the problems such as increasing sensitivity, specificity and accuracy. To overcome those problems we have to introduce an effective image processing algorithms for increasing performances and also easily identify the DR diseases. One of the most challenging tasks in screening is automatic detection of Microaneurysms (MAs). This paper presents a new approach to detect MAs. Our proposed work consists of preprocessing, blood vessel segmentation (FPCM), fovea localization, fovea elimination, feature extraction and classification (Neuro-Fuzzy). Neuro-Fuzzy is a combined version of neural networks and fuzzy logical models. Experiments are conducted using MATLAB simulation tool. Using MESSIDOR database for our experiments which provides efficient and effective results in sensitivity, specificity, correct classification and detection rate (accuracy) and precision.

Comparison of Optical Coherence Tomography Angiography and Fluorescein Angiography for the Identification of Retinal Vascular Changes in Eyes With Diabetic Macular Edema.

To compare the visualization of microaneurysms (MA) and the foveal avascular zone (FAZ) area using optical coherence tomography angiography (OCTA) versus fluorescein angiography (FA) in patients with diabetic macular edema (DME). Patients were prospectively recruited for same-day imaging on spectral-domain OCTA and FA. OCTA images were automatically segmented into superficial (sOCTA) and deep (dOCTA) capillary plexuses. The number of visible MAs and the FAZ area were compared between the two imaging modalities. Nineteen eyes of 10 patients were included. There was a statistically significant difference between MA counts for FA, sOCTA, and dOCTA (P = .002), and median MA counts were 14.5 (range: 2-43), 9.75 (range: 0-37.5), and 22.5 (range: 5.5-46.5), respectively. dOCTA showed significantly more MAs than sOCTA (P < .001). Although not significant statistically, dOCTA revealed more MAs than FA (P = .06). There was a statistically significant difference between FAZ area for FA, sOCTA, and dOCTA (P = .046), and median FAZ areas were 0.444 (range: 0.1-0.689), 0.224 (range: 0.335-0.806), and 0.345 (range: 0.106-0.881), respectively. FA showed a significantly larger FAZ area than sOCTA (P = .04). Deep plexus OCTA can better identify microaneurysms compared to either sOCTA or FA. The FAZ area appears larger on FA in contrast to OCTA of both plexuses. [Ophthalmic Surg Lasers Imaging Retina. 2016;47:1013-1019.].

Vessel segmentation and microaneurysm detection using discriminative dictionary learning and sparse representation

Diabetic retinopathy (DR) is a major cause of visual impairment, and the analysis of retinal image can assist patients to take action earlier when it is more likely to be effective. The accurate segmentation of blood vessels in the retinal image can diagnose DR directly. In this paper, a novel scheme for blood vessel segmentation based on discriminative dictionary learning (DDL) and sparse representation has been proposed. The proposed system yields a strong representation which contains the semantic concept of the image. To extract blood vessel, two separate dictionaries, for vessel and non-vessel, capable of providing reconstructive and discriminative information of the retinal image are learned. In the test step, an unseen retinal image is divided into overlapping patches and classified to vessel and non-vessel patches. Then, a voting scheme is applied to generate the binary vessel map. The proposed vessel segmentation method can achieve the accuracy of 95% and a sensitivity of 75% in the same range of specificity 97% on two public datasets. The results show that the proposed method can achieve comparable results to existing methods and decrease false positive vessels in abnormal retinal images with pathological regions. Microaneurysm (MA) is the earliest sign of DR that appears as a small red dot on the surface of the retina. Despite several attempts to develop automated MA detection systems, it is still a challenging problem. In this paper, a method for MA detection, which is similar to our vessel segmentation approach, is proposed. In our method, a candidate detection algorithm based on the Morlet wavelet is applied to identify all possible MA candidates. In the next step, two discriminative dictionaries with the ability to distinguish MA from non-MA object are learned. These dictionaries are then used to classify the detected candidate objects. The evaluations indicate that the proposed MA detection method achieves higher average sensitivity about 2–15%, compared to existing methods.

Effect of leaking perifoveal microaneurysms on resolution of diabetic macular edema treated by combination therapy using anti-vascular endothelial growth factor and short pulse focal/grid laser photocoagulation.

The effect of combination therapy using intravitreal ranibizumab (IVR) injections and short pulse focal/grid laser photocoagulation was evaluated for the treatment of diabetic macular edema (DME). The current investigation was a preliminary single-arm, open-label, prospective clinical study conducted on 21 eyes at 4 sites in Japan. Treatment protocol consisted of two phases. The induction IVR phase included two monthly IVRs followed by PRN IVR phase in which additional IVR was administered if the central macular thickness (CMT) exceeded 300μm. One week after each IVR in both phases, short pulse focal/grid laser was delivered to treat residual leakage outside of the fovea (>500μm) and reduce edema fluid influx. At the 6-month endpoint, the effects of treatment were examined in terms of best corrected visual acuity (BCVA), CMT, and required number of IVR injections in eyes with or without perifoveal leaking microaneurysms (MAs). In eyes with initial BCVA ≤70 letters, mean BCVA was significantly ameliorated by 7.0±7.4 letters (P=0.0324) and mean CMT improved significantly by 174.8±105.0µm (P=0.0005). Both BCVA improvement (P=0.8693) and CMT reduction (P=0.9336) were comparable between MA(-) and MA(+) groups. The MA(-) group required significantly fewer PRN-IVR injections than did the MA(+) group over the 6-month study period (mean 3.4±1.6 vs. 5.3±0.9, median 3.0 vs. 5.5; P=0.0229). Short pulse focal/grid laser photocoagulation could reduce the number of IVR injections required to resolve macular edema and increase BCVA in a possible mechanism of reduced influx of edema fluid into the foveal area in eyes without apparent perifoveal microaneurysms.

Noninvasive detection of microaneurysms in diabetic retinopathy by swept-source optical coherence tomography.

Background and objectivesA method of identifying retinal vascular microaneurysms (MAs) in nonproliferative diabetic retinopathy (NPDR) using swept-source optical coherence tomography (SS-OCT).Patients and methodsSS-OCT images were acquired in 17 eyes with NPDR using prototype SS-OCT device and fluorescein angiography (FA) images were obtained simultaneously. MAs identified on SS-OCT slabs were correlated to MAs identified on FA.ResultsMAs were identified in SS-OCT slabs in 15/17 eyes, resulting in NPDR diagnosis rate of 88%. Mean number of MAs identified on FA was 11.7±11.9 (total 199) and was 8.1±9.3 (total 137) on SS-OCT. Wilcoxon rank sum test showed no significant difference in MAs detected on SS-OCT and FA (P=0.2995) across eyes. Wilcoxon rank sum test showed SS-OCT detected slightly fewer MAs than FA per eye (3.65 less, P=0.0009).ConclusionSS-OCT visualization of MAs could serve as a tool for diagnosing NPDR, and possibly applied as an imaging biomarker for population-based diabetic retinopathy screening.

Can automated analysis of sequential retinal images of people attending diabetic retinopathy screening predict future referral to ophthalmology?

Introduction Providing systemic screening to a growing diabetes population is a challenge for most screening programmes. We have previously developed software for the automated detection of diabetic retinopathy which is now routinely used within the Scottish Diabetic Retinopathy Screening programme. Purpose In this study we explored whether automated analysis of sequential retinal images of people attending screening can predict future referral to ophthalmology. Materials and methods We developed software to automatically measure whether the same microaneurysm (MA), the main indicator of retinopathy, appeared in sequential images and whether new MAs appeared, how close the MAs were to the fovea, the number of MAs within each quadrant of the retina, and the presence of other indicators of retinopathy, namely exudates and haemorrhages. A retrospective cohort study was conducted using 12,754 subjects to assess whether these features predicted retinopathy development. Results A number of the novel features were independently associated with retinopathy progression of retinal images. These were higher MA counts close to the fovea, higher MA turnover between screening episodes and higher probabilities of haemorrhages anywhere in the image and exudates close to the fovea. We developed a model to estimate the risk of progression over the next 15 months. The sensitivity, specificity, PPV and NPV were 83.6%, 79.5%, 8.5% and 99.5%, respectively. We also developed a model for the risk of progression over the next 5 years. Conclusion Features of retinopathy derived from automated analysis of sequential photographs can help predict the risk of progression to retinopathy needing referral to ophthalmology.