Self-CephaloNet: a two-stage novel framework using operational neural network for cephalometric analysis

Neural Computing and Applications, Journal Year: 2025, Volume and Issue: unknown

Published: March 7, 2025

Abstract Cephalometric analysis is essential for the diagnosis and treatment planning of orthodontics. In lateral cephalograms, however, manual detection anatomical landmarks a time-consuming procedure. Deep learning solutions hold potential to address time constraints associated with certain tasks; concerns regarding their performances have been observed. To this critical issue, we propose an end-to-end cascaded deep framework (Self-CephaloNet) task, which demonstrates benchmark performance over ISBI 2015 dataset in predicting 19 cephalometric landmarks. Due adaptive nodal capabilities, Self-ONN (self-operational neural networks) superior complex feature spaces conventional convolutional networks. leverage attribute, introduce novel self-bottleneck HRNetV2 (high-resolution network) backbone, has exhibited on our landmark task. Our first-stage result surpasses previous studies, showcasing efficacy singular model, achieves remarkable 70.95% success rate detecting within 2-mm range Test1 Test2 datasets are part dataset. Moreover, second stage significantly improves overall performance, yielding impressive 82.25% average above same distance. Furthermore, external validation conducted using PKU cephalogram model commendable 75.95% range.

Language: Английский

---

Lightweight Evolving U-Net for Next-Generation Biomedical Imaging

Diagnostics, Journal Year: 2025, Volume and Issue: 15(9), P. 1120 - 1120

Published: April 28, 2025

Background/Objectives: Accurate and efficient segmentation of cell nuclei in biomedical images is critical for a wide range clinical research applications, including cancer diagnostics, histopathological analysis, therapeutic monitoring. Although U-Net its variants have achieved notable success medical image segmentation, challenges persist balancing accuracy with computational efficiency, especially when dealing large-scale datasets resource-limited settings. This study aims to develop lightweight scalable U-Net-based architecture that enhances performance while substantially reducing overhead. Methods: We propose novel evolving integrates multi-scale feature extraction, depthwise separable convolutions, residual connections, attention mechanisms improve robustness across diverse imaging conditions. Additionally, we incorporate channel reduction expansion strategies inspired by ShuffleNet minimize model parameters without sacrificing precision. The was extensively validated using the 2018 Data Science Bowl dataset. Results: Experimental evaluation demonstrates proposed achieves Dice Similarity Coefficient (DSC) 0.95 an 0.94, surpassing state-of-the-art benchmarks. effectively delineates complex overlapping structures high fidelity, maintaining efficiency suitable real-time applications. Conclusions: variant offers adaptable solution tasks. Its strong both highlights potential deployment diagnostics biological research, paving way resource-conscious solutions.

Language: Английский

---

Deep learning-based efficient diagnosis of periapical diseases with dental X-rays

Image and Vision Computing, Journal Year: 2024, Volume and Issue: 147, P. 105061 - 105061

Published: May 8, 2024

Language: Английский

---

A review of Artificial Intelligence methods in bladder cancer: segmentation, classification, and detection

Artificial Intelligence Review, Journal Year: 2024, Volume and Issue: 57(12)

Published: Oct. 21, 2024

Abstract Artificial intelligence (AI) and other disruptive technologies can potentially improve healthcare across various disciplines. Its subclasses, artificial neural networks, deep learning, machine excel in extracting insights from large datasets improving predictive models to boost their utility accuracy. Though research this area is still its early phases, it holds enormous potential for the diagnosis, prognosis, treatment of urological diseases, such as bladder cancer. The long-used nomograms classic forecasting approaches are being reconsidered considering AI’s capabilities. This review emphasizes coming integration into settings while critically examining most recent significant literature on subject. study seeks define status AI future, with a special emphasis how transform cancer diagnosis treatment.

Language: Английский

---

Label credibility correction based on cell morphological differences for cervical cells classification

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Jan. 2, 2025

Cervical cancer is one of the deadliest cancers that pose a significant threat to women's health. Early detection and treatment are commonly used methods prevent cervical cancer. The use pathological image analysis techniques for automatic interpretation cells in slides prominent area research field digital medicine. According Bethesda System, cytology necessitates further classification precancerous lesions based on positive interpretations. However, clinical definitions among different categories lesion complex often characterized by fuzzy boundaries. In addition, pathologists can deduce criteria judgment leading potential confusion during data labeling. Noisy labels due this reason great challenge supervised learning. To address problem caused noisy labels, we propose method label credibility correction cell images network. Firstly, contrastive learning network extract discriminative features from obtain more similar intra-class sample features. Subsequently, these fed into an unsupervised clustering, resulting class labels. Then corresponded true separate confusable typical samples. Through similarity comparison between cluster samples statistical feature centers each class, carried out group Finally, multi-class trained using synergistic grouping method. order enhance stability model, momentum incorporated loss. Experimental validation conducted dataset comprising approximately 60,000 multiple hospitals, showcasing effectiveness our proposed approach. achieves 2-class task accuracy 0.9241 5-class 0.8598. Our better performance than existing networks

Language: Английский

---

A Cascade Model to Detect and Segment Lung Nodule Using YOLOv8 and Resnet50U‐Net

International Journal of Imaging Systems and Technology, Journal Year: 2025, Volume and Issue: 35(1)

Published: Jan. 1, 2025

ABSTRACT In our research, we introduce a sophisticated “two‐stage” or cascade model designed to enhance the precision of lung nodule analysis. This innovative approach integrates two crucial processes: detection and segmentation. initial stage, specialized object algorithm efficiently scans medical images identify potential areas interest, specifically focusing on nodules. plays role in minimizing segmentation area, particularly context imaging, where structures exhibit heterogeneity. helps focus process only relevant areas, reducing unnecessary computation errors. Subsequently, second stage employs advanced algorithms precisely delineate boundaries identified nodules, providing detailed accurate contours. The combination not enhances overall accuracy cancer but also minimizes false positives, streamlines workflow for radiologists, provides more comprehensive understanding abnormalities. Additionally, it improves efficiency segmentation, especially cases complexity heterogeneity structure make task challenging. proposed method has been tested LIDC‐IDRI dataset, demonstrating favorable results both steps, with 81.3% mAP 83.54% DSC, respectively. These serve as evidence that effectively

Language: Английский

---

Multi-class Classification of Retinal Eye Diseases from Ophthalmoscopy Images Using Transfer Learning-Based Vision Transformers

Deleted Journal, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 27, 2025

This study explores a transfer learning approach with vision transformers (ViTs) and convolutional neural networks (CNNs) for classifying retinal diseases, specifically diabetic retinopathy, glaucoma, cataracts, from ophthalmoscopy images. Using balanced subset of 4217 images ophthalmology-specific pretrained ViT backbones, this method demonstrates significant improvements in classification accuracy, offering potential broader applications medical imaging. Glaucoma, cataracts are common eye diseases that can cause loss if not treated. These must be identified the early stages to prevent damage progression. paper focuses on accurate identification analysis disparate including using Deep (DL) has been widely used image recognition detection treatment diseases. In study, ResNet50, DenseNet121, Inception-ResNetV2, six variations employed, their performance diagnosing such as retinopathy is evaluated. particular, article uses transformer model an automated diagnose highlighting accuracy pre-trained deep (DTL) structures. The updated ViT#5 augmented-regularized (AugReg ViT-L/16_224) rate 0.00002 outperforms state-of-the-art techniques, obtaining data-based score 98.1% publicly accessible dataset, which includes most categories, other convolutional-based models terms precision, recall, F1 score. research contributes significantly analysis, demonstrating AI enhancing precision disease diagnoses advocating integration artificial intelligence diagnostics.

Language: Английский

---

Automated breast nuclei feature extraction for segmentation in histopathology images using Deep-CNN-based gaussian mixture model and color optimization technique

Multimedia Tools and Applications, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 18, 2025

Language: Английский

---

Semi-supervised tissue segmentation from histopathological images with consistency regularization and uncertainty estimation

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 22, 2025

Abstract Pathologists have depended on their visual experience to assess tissue structures in smear images, which was time-consuming, error-prone, and inconsistent. Deep learning, particularly Convolutional Neural Networks (CNNs), offers the ability automate this procedure by recognizing patterns images. However, training these models necessitates huge amounts of labeled data, can be difficult come due skill required for annotation unavailability rare diseases. This work introduces a new semi-supervised method structure semantic segmentation histopathological The study presents CNN based teacher model that generates pseudo-labels train student model, aiming overcome drawbacks conventional supervised learning approaches. Self-supervised is used improve model’s performance smaller datasets. Consistency regularization integrated efficiently data. Further, uses Monte Carlo dropout estimate uncertainty proposed model. demonstrated promising results achieving an mIoU score 0.64 public dataset, highlighting its potential accuracy image analysis.

Language: Английский

---

MCU-Net: A multi-prior collaborative deep unfolding network with gates-controlled spatial attention for accelerated MRI reconstruction

Neurocomputing, Journal Year: 2025, Volume and Issue: 633, P. 129771 - 129771

Published: Feb. 26, 2025

Language: Английский

---