Medical images classification using deep learning: a survey

Multimedia Tools and Applications, Journal Year: 2023, Volume and Issue: 83(7), P. 19683 - 19728

Published: July 28, 2023

Language: Английский

---

Deep Learning and Machine Learning Algorithms for Retinal Image Analysis in Neurodegenerative Disease: Systematic Review of Datasets and Models

Translational Vision Science & Technology, Journal Year: 2024, Volume and Issue: 13(2), P. 16 - 16

Published: Feb. 21, 2024

Purpose: Retinal images contain rich biomarker information for neurodegenerative disease. Recently, deep learning models have been used automated disease diagnosis and risk prediction using retinal with good results. Methods: In this review, we systematically report studies datasets of from patients diseases, including Alzheimer's disease, Huntington's Parkinson's amyotrophic lateral sclerosis, others. We also review characterize the in current literature which classification, regression, or segmentation problems diseases. Results: Our found several existing various imaging modalities primarily most on order tens to a few hundred images. limited data available other Although cross-sectional is becoming more abundant, longitudinal any are lacking. Conclusions: The use bilateral multimodal together metadata seems improve model performance, thus image patient needed. identified tools that useful context feature extraction algorithms specifically images, preprocessing techniques, transfer learning, fusion, attention mapping. Importantly, consider limitations common these real-world clinical applications. Translational Relevance: This systematic evaluates features relevant evaluation

Language: Английский

---

RetinaRegNet: A zero-shot approach for retinal image registration

Computers in Biology and Medicine, Journal Year: 2025, Volume and Issue: 186, P. 109645 - 109645

Published: Jan. 14, 2025

Retinal image registration is essential for monitoring eye diseases and planning treatments, yet it remains challenging due to large deformations, minimal overlap, varying quality. To address these challenges, we propose RetinaRegNet, a multi-stage model with zero-shot generalizability across multiple retinal imaging modalities. RetinaRegNet begins by extracting features using pretrained latent diffusion model. Feature points are sampled from the fixed combination of SIFT algorithm random sampling. For each point, its corresponding point in moving estimated cosine similarities between feature vectors that all pixels image. Outliers correspondences detected an inverse consistency constraint, ensuring both forward backward directions. distances true further removed transformation-based outlier detector. The resulting then used estimate geometric transformation two images. We use two-stage framework robust accurate alignment: first stage estimates homography global alignment, second third-order polynomial capture local deformations. evaluated on three modalities: color fundus, fluorescein angiography, laser speckle flowgraphy. Across datasets, consistently outperformed state-of-the-art methods, achieving AUC scores 0.901, 0.868, 0.861, respectively. RetinaRegNet's performance highlights potential as valuable tool tracking disease progression evaluating treatment efficacy. Our code publicly available at: https://github.com/mirthAI/RetinaRegNet.

Language: Английский

---

An extensive review of state-of-the-art transfer learning techniques used in medical imaging: Open issues and challenges

Journal of Intelligent Systems, Journal Year: 2022, Volume and Issue: 31(1), P. 1085 - 1111

Published: Jan. 1, 2022

Abstract Deep learning techniques, which use a massive technology known as convolutional neural networks, have shown excellent results in variety of areas, including image processing and interpretation. However, the depth these networks grows, so does demand for large amount labeled data required to train networks. In particular, medical field suffers from lack images because procedure obtaining healthcare is difficult, expensive, requires specialized expertise add labels images. Moreover, process may be prone errors time-consuming. Current research has revealed transfer viable solution this problem. Transfer allows us knowledge gained previous improve tackle new This study aims conduct comprehensive survey recent studies that dealt with solving problem most important metrics used evaluate methods. addition, identifies problems techniques highlights dataset potential can addressed future research. According our review, many researchers pre-trained models on Imagenet (VGG16, ResNet, Inception v3) applications such skin cancer, breast diabetic retinopathy classification tasks. These require further investigation models, due training them natural, non-medical augmentation expand their avoid overfitting. not enough effect performance or without augmentation. Accuracy, recall, precision, F 1 score, receiver operator characteristic curve, area under curve (AUC) were widely measures studies. Furthermore, we identified datasets melanoma cancer suggested corresponding solutions.

Language: Английский

---

VisionDeep-AI: Deep learning-based retinal blood vessels segmentation and multi-class classification framework for eye diagnosis

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 94, P. 106273 - 106273

Published: March 28, 2024

Language: Английский

---

A three-stage novel framework for efficient and automatic glaucoma classification from retinal fundus images

Multimedia Tools and Applications, Journal Year: 2024, Volume and Issue: unknown

Published: June 14, 2024

Language: Английский

---

ResMU-Net: Residual Multi-kernel U-Net for blood vessel segmentation in retinal fundus images

Biomedical Signal Processing and Control, Journal Year: 2023, Volume and Issue: 90, P. 105859 - 105859

Published: Dec. 17, 2023

Language: Английский

---

HWA-SegNet: Multi-channel skin lesion image segmentation network with hierarchical analysis and weight adjustment

Computers in Biology and Medicine, Journal Year: 2022, Volume and Issue: 152, P. 106343 - 106343

Published: Nov. 28, 2022

Language: Английский

---

GCHA-Net: Global context and hybrid attention network for automatic liver segmentation

Computers in Biology and Medicine, Journal Year: 2022, Volume and Issue: 152, P. 106352 - 106352

Published: Nov. 26, 2022

Language: Английский

---

Grad-CAM based explanations for multiocular disease detection using Xception net

Image and Vision Computing, Journal Year: 2025, Volume and Issue: unknown, P. 105419 - 105419

Published: Jan. 1, 2025

Language: Английский

---