DRDrV3: Complete Lesion Detection in Fundus Images Using Mask R-CNN, Transfer Learning, and LSTM

Shenavarmasouleh, Farzan; Mohammadi, Farid Ghareh; Amini, M. Hadi; Taha, Thiab; Rasheed, Khaled; Arabnia, Hamid R.

doi:10.48550/arxiv.2108.08095

Cited by 2 publications

(6 citation statements)

References 21 publications

(35 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The results were achieved in the DDR validation set using the metrics AP and mAP with a limit of IoU of 0.5. The best mAP result obtained by the proposed approach in the lesion detection task (BBox) was 0.2903 (indicated in bold), which was lower than the result presented by the work by Shenavarmasouleh et al [5], which presented an mAP of 0.4780 (shown in bold). However, it should be noted that Shenavarmasouleh et al [5] was limited to detecting only two classes of lesions: Exudates and Microaneurysms.…”

Section: Table 12contrasting

confidence: 65%

“…Therefore, in future work, the authors state that additional validation is needed through public datasets to properly assess the system's performance in DR classification and the detection of fundus lesions. Shenavarmasouleh and Arabnia [14] and Shenavarmasouleh et al [5] propose an architecture for detecting fundus lesions composed of two modules: a module for detecting lesions and another module for classifying the severity of the DR. The proposed solution was evaluated using the metrics IoU and mAP.…”

Section: Related Workmentioning

confidence: 99%

“…In the first stage, regions of interest (ROI) are generated by an RPN module; in the second stage, the class outputs, bounding box regression, and binary masks are generated using the ROI generated in the first stage. For each region of interest analyzed, a multitasking loss function [33] is applied, according to Equation (5): Loss = L Class + L BBox + L Mask (5) where L Class corresponds to classification loss, L BBox corresponds to bounding box regression loss, and L Mask corresponds to mask loss. The result of the final loss function is obtained based on the average loss over the samples.…”

Section: Deep Neural Network Architecturementioning

confidence: 99%

“…A computerized approach can aid in the process of identifying retinal lesions and aid in the diagnosis and screening of DR. Solutions based on deep neural networks have been presented in the literature to help identify the disease, such as the works proposed by Porwal et al [2], Li et al [3], Dai et al [4] and Shenavarmasouleh et al [5]. Although these works have shown promising results, there are still limitations in the results presented, which are mainly associated with identifying microlesions in retinal images.…”

mentioning

confidence: 99%

See 3 more Smart Citations

A New Approach for Fundus Lesions Instance Segmentation Based on Mask R-CNN X101-FPN Pre-Trained Architecture

et al. 2023

View full text Add to dashboard Cite

Diabetic retinopathy is one of the main causes of vision loss, and it can be identified through ophthalmological examinations that aim to locate the presence of retinal lesions such as Microaneurysms, Hemorrhages, Soft Exudates, and Hard Exudates. The development of computerized approaches to perform the instance segmentation of these lesions can help in the early diagnosis of the disease. However, the segmentation of instances of artifacts in the retina is a complex task due to factors such as object size and morphological characteristics. This article proposes a new approach based on a Mask Regions with Convolutional Neural Network features (Mask R-CNN) architecture to perform instance segmentation of lesions associated with diabetic retinopathy. The proposed approach was trained, adjusted, and tested using different public datasets of diabetic retinopathy, which were implemented with the Detectron2 libraries and OpenCV. The best result obtained by the proposed approach in the Dataset for Diabetic Retinopathy (DDR) was using Tilling and Adam optimizer, reaching a mean Average Precision (mAP) of 0.2903 in the detection of fundus lesions for the limit of Intersection Over Union (IoU ) of 0.5 in the validation stage and an mAP of 0.1670 in the detection of fundus lesions to the limit of IoU of 0.5 in the test step. The results obtained in the experiments demonstrate that the proposed approach presented promising results in the instance segmentation of microaneurysms with an increase in precision, which in our case reaches approximately 16%.

show abstract

Section: Table 12contrasting

confidence: 65%

Section: Related Workmentioning

confidence: 99%

Section: Deep Neural Network Architecturementioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

A New Approach for Fundus Lesions Instance Segmentation Based on Mask R-CNN X101-FPN Pre-Trained Architecture

et al. 2023

View full text Add to dashboard Cite

show abstract

“…Apart from playing a pivotal role in the recent advances of the main fields, these datasets also proved to be useful when used with transfer learning methods to help underlying disciplines such as biomedical imaging [18,19,20]. However, the aforementioned datasets are prune to restrictions.…”

Section: Introductionmentioning

confidence: 99%

Embodied AI-Driven Operation of Smart Cities: A Concise Review

Shenavarmasouleh¹,

Mohammadi²,

Amini³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

A smart city can be seen as a framework, comprised of Information and Communication Technologies (ICT). An intelligent network of connected devices that collect data with their sensors and transmit them using wireless and cloud technologies in order to communicate with other assets in the ecosystem plays a pivotal role in this framework. Maximizing the quality of life of citizens, making better use of available resources, cutting costs, and improving sustainability are the ultimate goals that a smart city is after. Hence, data collected from these connected devices will continuously get thoroughly analyzed to gain better insights into the services that are being offered across the city; with this goal in mind that they can be used to make the whole system more efficient. Robots and physical machines are inseparable parts of a smart city. Embodied AI is the field of study that takes a deeper look into these and explores how they can fit into real-world environments. It focuses on learning through interaction with the surrounding environment, as opposed to Internet AI which tries to learn from static datasets. Embodied AI aims to train an agent that can See (Computer Vision), Talk (NLP), Navigate and Interact with its environment (Reinforcement Learning), and Reason (General Intelligence), all at the same time. Autonomous driving cars and personal companions are some of the examples that benefit from Embodied AI nowadays. In this paper, we attempt to do a concise review of this field. We will go through its definitions, its characteristics, and its current achievements along with different algorithms, approaches, and solutions that are being used in different components of it (e.g. Vision, NLP, RL). We will then explore all the available simulators and 3D interactable databases that will make the research in this area feasible. Finally, we will address its challenges and identify its potentials for future research.

show abstract

DRDrV3: Complete Lesion Detection in Fundus Images Using Mask R-CNN, Transfer Learning, and LSTM

Cited by 2 publications

References 21 publications

A New Approach for Fundus Lesions Instance Segmentation Based on Mask R-CNN X101-FPN Pre-Trained Architecture

A New Approach for Fundus Lesions Instance Segmentation Based on Mask R-CNN X101-FPN Pre-Trained Architecture

Embodied AI-Driven Operation of Smart Cities: A Concise Review

Contact Info

Product

Resources

About