Advances in Medical Image Segmentation: A Comprehensive Review of Traditional, Deep Learning and Hybrid Approaches

Bioengineering, Год журнала: 2024, Номер 11(10), С. 1034 - 1034

Опубликована: Окт. 16, 2024

Medical image segmentation plays a critical role in accurate diagnosis and treatment planning, enabling precise analysis across wide range of clinical tasks. This review begins by offering comprehensive overview traditional techniques, including thresholding, edge-based methods, region-based approaches, clustering, graph-based segmentation. While these methods are computationally efficient interpretable, they often face significant challenges when applied to complex, noisy, or variable medical images. The central focus this is the transformative impact deep learning on We delve into prominent architectures such as Convolutional Neural Networks (CNNs), Fully (FCNs), U-Net, Recurrent (RNNs), Adversarial (GANs), Autoencoders (AEs). Each architecture analyzed terms its structural foundation specific application segmentation, illustrating how models have enhanced accuracy various contexts. Finally, examines integration with addressing limitations both approaches. These hybrid strategies offer improved performance, particularly challenging scenarios involving weak edges, noise, inconsistent intensities. By synthesizing recent advancements, provides detailed resource for researchers practitioners, valuable insights current landscape future directions

Язык: Английский

---

A priority-guided contrastive network for delineating vascular layers in arterial ultrasound

Expert Systems with Applications, Год журнала: 2025, Номер 272, С. 126695 - 126695

Опубликована: Фев. 3, 2025

Язык: Английский

---

Implementation of Chatbot that Predicts an Illness Dynamically using Machine Learning Techniques

International journal of engineering. Transactions B: Applications, Год журнала: 2023, Номер 37(2), С. 312 - 322

Опубликована: Ноя. 28, 2023

Timely access to healthcare is crucial in order maintain a high standard of living. However, obtaining medical consultations can be difficult, especially for those living remote areas or during pandemic when face-to-face are not always possible. The ability accurately diagnose diseases essential effective treatment, and recent technological advancements offer potential solution. Machine learning (ML) Natural language processing (NLP) enables computer programs understand human extract desired features from responses, allowing human-like interaction with users. By leveraging these technologies, professionals potentially provide more accessible efficient individuals, regardless their location. concept establish an online platform where users ask medical-related queries receive responses both fellow would feature Medical Chatbot, which employs advanced ML techniques analyze user-provided symptoms initial disease diagnosis related information prior consulting doctor. This prediction chatbot interacts dynamically the enter based on syntactic semantic similarity response given. In this work threshold score kept 0.7. K-Nearest neighbors, Random forest, Support vector machine, Naive bayes Logistic regression algorithms used faced by similarity, fuzzy string matching using all-MiniLM-L6-v2 model improve efficiency result.

Язык: Английский

---

Traffic Scene Analysis and Classification using Deep Learning

International journal of engineering. Transactions C: Aspects, Год журнала: 2023, Номер 37(3), С. 496 - 502

Опубликована: Дек. 23, 2023

In this study, we aim to use new deep-learning tools and convolutional neural networks for traffic analysis. ResNeXt architecture, one of the most potent architectures has attracted much attention in various fields, been proposed examine scene, classify it into three categories: cars, bikes (bicycles/motorcycles), pedestrians. Previous studies have focused more on type classification reported only human-facial recognition or vehicle detection. contrast, method uses precise architecture perform classes. The plan implemented several steps: first stage is divide critical objects. next step, characteristics obtained objects are extracted process Experiments conducted different essential datasets such as high-traffic, low-quality, real-time scenes. Essential evaluation criteria accuracy, sensitivity, specificity show that performance improved compared methods being compared. accuracy criterion reached than 92%, sensitivity about 89%, specially 90.25%. can be used implement intelligent cities, public safety, metropolitan decisions results urban management, predictive modeling lost data sequential generalizability.

Язык: Английский

---

Convolutional Neural Network Application for Detection & Classification of Brain Tumour

2018 5th IEEE Uttar Pradesh Section International Conference on Electrical, Electronics and Computer Engineering (UPCON), Год журнала: 2023, Номер unknown, С. 1482 - 1486

Опубликована: Дек. 1, 2023

Brain tumours may be either benign or malignant. The highest grade of brain is associated with a very poor survival rate. So, plan your treatments ahead time. Improve the patients' living conditions on stage. Cancers brain, lung, liver, breast, and prostate are often evaluated using imaging modalities such computed tomography (CT), magnetic resonance (MRI), ultrasound images. In this research, MRI images employed specifically for diagnosis cancer. Unfortunately, it currently difficult to manually categorise tumour from non-tumour scan due sheer volume data generated by scans. There limitation, too, in that only limited number can reliably get quantitative information. Thus, reducing death rate among humans depends critically trustworthy automated categorization system. notoriously automatically classify wide variety locations surrounding tissues. authors propose CNN classification quickly easily identify cancers. underlying architecture developed small kernels. has been great deal research towards improving efficiency which different kinds identified. Segmenting, identifying, extracting contaminated region (MR) time-consuming labour-intensive process relies heavily expertise clinician doing procedure. Because crucial use computer-aided technologies. We evaluate size Convolutional Neural Network method, consistently yields accurate results.

Язык: Английский

---

Improving Deep Learning-based Saliency Detection Using Channel Attention Module

International journal of engineering. Transactions B: Applications, Год журнала: 2024, Номер 37(11), С. 2367 - 2379

Опубликована: Янв. 1, 2024

In recent decades, the advancement of deep learning algorithms and their effectiveness in saliency detection has garnered significant attention research. Among these methods, U Network ( U-Net ) is widely used computer vision image processing. However, most previous learning-based methods have focused on accuracy salient regions, often overlooking quality boundaries, especially fine boundaries. To address this gap, we developed a method to detect boundaries effectively. This comprises two modules: prediction residual refinement, based structure. The refinement module improves mask predicted by module. Additionally, boost map, channel integrated. impact our proposed method. implemented module, aiding network obtaining more accurate estimation focusing crucial informative regions image. evaluate method, five well-known datasets are employed. consistently outperforms baseline across all datasets, demonstrating improved performance.

Язык: Английский

---

Systematic Survey of Deep Fuzzy Computer Vision in Biomedical Research

Fuzzy Information and Engineering, Год журнала: 2024, Номер 16(3), С. 220 - 243

Опубликована: Сен. 1, 2024

Язык: Английский

---

Evaluation method of distribution network operation status based on local fuzzy measure in boundary region

Energy Informatics, Год журнала: 2024, Номер 7(1)

Опубликована: Ноя. 25, 2024

With the increasing complexity of distribution network, proportion abnormal data in monitoring network and its daily work is extremely low. Traditional clustering analysis methods are difficult to effectively solve imbalance problem. Therefore, this paper introduces influence parameters that can adaptively adjust cluster center local samples boundary area, improves update formula, proposes a method operation state evaluation based on blur measurement region. The research results found five indicators proposed algorithm were 112, 0, 2, 26, 5, respectively, all which superior comparison algorithms. showed optimization fuzzy measure region could reduce negative impact edge occupied by most clusters effect, so was always an ideal position. At same time, example had risk prediction 0.91 for power outage networks, close real situation high accuracy. It provide reference maintenance grid personnel, eliminate hidden dangers advance, ensure safe grid.

Язык: Английский

---

Algorithm of Predicting Heart Attack with using Sparse Coder

International journal of engineering. Transactions C: Aspects, Год журнала: 2023, Номер 36(12), С. 2190 - 2197

Опубликована: Янв. 1, 2023

One of the most serious causes disease in world's population, which kills many people worldwide every year, is heart attack. Various factors are involved this matter, such as high blood pressure, cholesterol, abnormal pulse rate, diabetes, etc. methods have been proposed field, but article, by using sparse codes classification process, higher accuracy has achieved predicting attacks. The method consists two parts: preprocessing and code processing. resistant to noise data scattering because it uses a representation for purpose. spars allow signal be displayed at its lowest value, leads improve computing speed reduce storage requirements. To evaluate method, Cleveland database used, includes 303 samples each sample 76 features. Only 13 features used method. FISTA, AMP, DALM PALM classifiers process. especially with classifier, highest among other 96.23%, 95.08%, 94.11% 94.52% DALM, respectively.

Язык: Английский

---

Survey of Brain Tumor Image Segmentation Using Artificial Intelligence Techniques

International Research Journal of Innovations in Engineering and Technology, Год журнала: 2023, Номер 07(12), С. 2581 - 3048

Опубликована: Янв. 1, 2023

A brain tumor is an abnormal tissue mass resulting from cell growth.Brain tumors often reduce the length of a person's life and may cause death in advanced cases.Physician teams rely on early detection accurate placement by magnetic resonance imaging to assess tumor's pace accuracy.Treatment, as well determining causes injury cells, further aids reducing any potential problems patient could experience.Segmenting images taken important for neurosurgeons.It not easy matter requires high experience radiologists.Therefore, there need expert intelligent system segment part medication that expert, designed address this task.One most promising innovative approaches medical industry artificial intelligence.Automatically identifying aberrant region made possible application intelligence imaging, which dependent picture interpretation.The goal research provide brief survey automatic methods segmentation using methods, includes use machine learning deep include several including (CNN, RES NET, MOBILE NET etc) are applied field, identify obtain results images.

Язык: Английский

---