A Generalised Vision Transformer-Based Self-Supervised Model for Diagnosing and Grading Prostate Cancer Using Histological Images DOI Creative Commons
Abadh K. Chaurasia, H Harris,

Patrick W Toohey

et al.

Published: Dec. 5, 2024

BACKGROUND: Gleason grading remains the gold standard for prostate cancer histological classification and prognosis, yet its subjectivity leads to grade variability between pathologists, potentially impacting clinical decision-making. Herein, we trained validated a generalised AI-driven system diagnosing using diverse datasets from tissue microarray (TMA) core whole slide images (WSIs) with Hematoxylin Eosin staining. METHODS: We analysed eight datasets, which included 12,711 3,648 patients, incorporating TMA WSIs. The Macenko method was used normalise colours consistency across images. Subsequently, multi-resolution (5x, 10x, 20x, 40x) binary classifier identify benign malignant tissue. then implemented multi-class patterns (GP) sub-categorisation Finally, models were externally on 11,132 histology 2,176 patients determine International Society of Urological Pathology (ISUP) grade. Models assessed various metrics, agreement model’s predictions ground truth quantified quadratic weighted Cohen’s Kappa (_κ_) score. RESULTS: Our demonstrated robust performance in distinguishing _κ_ scores 0.967 internal validation. model achieved ranging 0.876 0.995 four unseen testing datasets. also distinguished GP3, GP4, GPs an overall score 0.841. This further tested obtaining 0.774 0.888. models’ compared against independent pathologist’s annotation external dataset, achieving 0.752 classes. CONCLUSION: self-supervised ViT-based effectively diagnoses grades images, tissues classifying malignancies by aggressiveness. External validation highlights robustness applicability digital pathology.

Language: Английский

AutYOLO-ATT: an attention-based YOLOv8 algorithm for early autism diagnosis through facial expression recognition DOI Creative Commons

Reham Hosney,

Fatma M. Talaat, Eman M. El-Gendy

et al.

Neural Computing and Applications, Journal Year: 2024, Volume and Issue: 36(27), P. 17199 - 17219

Published: June 6, 2024

Abstract Autism Spectrum Disorder (ASD) is a developmental condition resulting from abnormalities in brain structure and function, which can manifest as communication social interaction difficulties. Conventional methods for diagnosing ASD may not be effective the early stages of disorder. Hence, diagnosis crucial to improving patient's overall health well-being. One alternative method autism facial expression recognition since autistic children typically exhibit distinct expressions that aid distinguishing them other children. This paper provides deep convolutional neural network (DCNN)-based real-time emotion system kids. The proposed designed identify six emotions, including surprise, delight, sadness, fear, joy, natural, assist medical professionals families recognizing intervention. In this study, an attention-based YOLOv8 (AutYOLO-ATT) algorithm proposed, enhances model's performance by integrating attention mechanism. outperforms all classifiers metrics, achieving precision 93.97%, recall 97.5%, F1-score 92.99%, accuracy 97.2%. These results highlight potential real-world applications, particularly fields where high essential.

Language: Английский

Citations

6
A Comprehensive Review of AI Diagnosis Strategies for Age-Related Macular Degeneration (AMD) DOI Creative Commons

Aya A. Abd El-Khalek,

Hossam Magdy Balaha, Ashraf Sewelam

et al.

Bioengineering, Journal Year: 2024, Volume and Issue: 11(7), P. 711 - 711

Published: July 13, 2024

The rapid advancement of computational infrastructure has led to unprecedented growth in machine learning, deep and computer vision, fundamentally transforming the analysis retinal images. By utilizing a wide array visual cues extracted from fundus images, sophisticated artificial intelligence models have been developed diagnose various disorders. This paper concentrates on detection Age-Related Macular Degeneration (AMD), significant condition, by offering an exhaustive examination recent learning methodologies. Additionally, it discusses potential obstacles constraints associated with implementing this technology field ophthalmology. Through systematic review, research aims assess efficacy techniques discerning AMD different modalities as they shown promise disorders diagnosis. Organized around prevalent datasets imaging techniques, initially outlines assessment criteria, image preprocessing methodologies, frameworks before conducting thorough investigation diverse approaches for detection. Drawing insights more than 30 selected studies, conclusion underscores current trajectories, major challenges, future prospects diagnosis, providing valuable resource both scholars practitioners domain.

Language: Английский

Citations

4
Enhanced handwriting recognition through hybrid UNet-based architecture with global classical features DOI
Xiaofei Liu

Journal of Ambient Intelligence and Humanized Computing, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 18, 2025

Language: Английский

Citations

0
Automating Prostate Cancer Grading: A Novel Deep Learning Framework for Automatic Prostate Cancer Grade Assessment using Classification and Segmentation DOI
Saidul Kabir, Rusab Sarmun,

Rafif Mahmood Al Saady

et al.

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

Published: Feb. 6, 2025

Prostate Cancer (PCa) is the second most common cancer in men and affects more than a million people each year. Grading prostate based on Gleason grading system, subjective labor-intensive method for evaluating tissue samples. The variability diagnostic approaches underscores urgent need reliable methods. By integrating deep learning technologies developing automated systems, precision can be improved, human error minimized. present work introduces three-stage framework-based innovative deep-learning system assessing PCa severity using PANDA challenge dataset. After meticulous selection process, 2699 usable cases were narrowed down from initial 5160 after extensive data cleaning. There are three stages proposed framework: classification of grades neural networks (DNNs), segmentation grades, computation International Society Urological Pathology (ISUP) machine classifiers. Four classes patches classified segmented (benign, 3, 4, 5). Patch sampling at different sizes (500 × 500 1000 pixels) was used to optimize processes. performance network enhanced by Self-organized operational (Self-ONN) DeepLabV3 architecture. Based these predictions, distribution percentages grade within whole slide images (WSI) calculated. These features then concatenated into classifiers predict final ISUP grade. EfficientNet_b0 achieved highest F1-score 83.83% classification, while + architecture self-ONN EfficientNet encoder Dice Similarity Coefficient (DSC) score 84.9% segmentation. Using RandomForest (RF) classifier, framework quadratic weighted kappa (QWK) 0.9215. Deep frameworks being developed automatically have shown promising results. In addition, it provides prospective approach prognostic tool that produce clinically significant results efficiently reliably. Further investigations needed evaluate framework's adaptability effectiveness across various clinical scenarios.

Language: Английский

Citations

0
A generalised vision transformer-based self-supervised model for diagnosing and grading prostate cancer using histological images DOI Creative Commons
Abadh K. Chaurasia, H Harris,

Patrick W Toohey

et al.

Prostate Cancer and Prostatic Diseases, Journal Year: 2025, Volume and Issue: unknown

Published: March 14, 2025

Abstract Background Gleason grading remains the gold standard for prostate cancer histological classification and prognosis, yet its subjectivity leads to grade variability between pathologists, potentially impacting clinical decision-making. Herein, we trained validated a generalised AI-driven system diagnosing using diverse datasets from tissue microarray (TMA) core whole slide images (WSIs) with Haematoxylin Eosin staining. Methods We analysed eight datasets, which included 12,711 3648 patients, incorporating TMA WSIs. The Macenko method was used normalise colours consistency across images. Subsequently, multi-resolution (5x, 10x, 20x, 40x) binary classifier identify benign malignant tissue. then implemented multi-class patterns (GP) sub-categorisation Finally, models were externally on 11,132 histology 2176 patients determine International Society of Urological Pathology (ISUP) grade. Models assessed various metrics, agreement model’s predictions ground truth quantified quadratic weighted Cohen’s Kappa ( κ ) score. Results Our demonstrated robust performance in distinguishing scores 0.967 internal validation. model achieved ranging 0.876 0.995 four unseen testing datasets. also distinguished GP3, GP4, GPs an overall score 0.841. This further tested obtaining 0.774 0.888. models’ compared against independent pathologist’s annotation external dataset, achieving 0.752 classes. Conclusion self-supervised ViT-based effectively diagnoses grades images, tissues classifying malignancies by aggressiveness. External validation highlights robustness applicability digital pathology.

Language: Английский

Citations

0
Prostate cancer classification using adaptive swarm Intelligence based deep attention neural network DOI

D Sowmya,

Siriki Atchuta Bhavani,

V. V. S. Sasank

et al.

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 96, P. 106654 - 106654

Published: July 19, 2024

Language: Английский

Citations

1
Towards ovarian cancer diagnostics: A vision transformer-based computer-aided diagnosis framework with enhanced interpretability DOI Creative Commons
Abdulrahman Alahmadi

Results in Engineering, Journal Year: 2024, Volume and Issue: 23, P. 102651 - 102651

Published: Aug. 2, 2024

Ovarian cancer, a significant threat to women's health, demands innovative diagnostic approaches. This paper introduces groundbreaking Computer-Aided Diagnosis (CAD) framework for the classification of ovarian integrating Vision Transformer (ViT) models and Local Interpretable Model-agnostic Explanations (LIME). ViT models, including ViT-Base-P16-224-In21K, ViT-Base-P16-224, ViT-Base-P32-384, ViT-Large-P32-384, exhibit exceptional accuracy, precision, recall, overall robust performance across diverse evaluation metrics. The incorporation stacked model further enhances performance. Experimental results, conducted on UBC-OCEAN training testing datasets, highlight proficiency in accurately classifying cancer subtypes based histopathological images. ViT-Large-P32-384 stands out as top performer, achieving 98.79% accuracy during 97.37% testing. Visualizations, Receiver Operating Characteristic (ROC) curves (LIME), provide insights into discriminative capabilities enhance interpretability. proposed CAD represents advancement diagnostics, offering promising avenue accurate transparent multi-class

Language: Английский

Citations

1
Optimization of Hydropower Unit Startup Process Based on the Improved Multi-Objective Particle Swarm Optimization Algorithm DOI Creative Commons
Qingquan Zhang,

Zifeng Xie,

Mingming Lu

et al.

Energies, Journal Year: 2024, Volume and Issue: 17(17), P. 4473 - 4473

Published: Sept. 6, 2024

In order to improve the dynamic performance during startup process of hydropower units, while considering efficient and stable speed increase effective suppression water pressure fluctuations mechanical vibrations, optimization algorithms must be used select optimal parameters for system. However, in current research, various multi-objective still have limitations terms target space coverage diversity maintenance parameter hydraulic turbines. To explore verify turbines, multiple strategies are proposed this study. Under condition constructing a fine-tuned nonlinear model control system, paper focuses on three key indicators: absolute integral deviation, snail shell fluctuation, relative value maximum axial thrust. Through comparative analysis particle swarm algorithm (MOPSO), variant (VMOPSO), sine cosine (MOSCA), biogeography (MOBBO), gravity search (MOGAS), improved (IMOPSO), obtained compared analyzed strategy, most suitable actual working conditions selected through comprehensive weighting method. The results show that, local solution problem caused by other algorithms, method significantly reduces vibrations ensuring improvement, achieving better performance. significant guiding significance smooth operation safety provide strong support making operational decisions.

Language: Английский

Citations

1
Advancing eye disease detection: A comprehensive study on computer-aided diagnosis with vision transformers and SHAP explainability techniques DOI
Hossam Magdy Balaha, Asmaa El-Sayed Hassan,

Ranaa Ahmed

et al.

Journal of Applied Biomedicine, Journal Year: 2024, Volume and Issue: 45(1), P. 23 - 33

Published: Dec. 15, 2024

Language: Английский

Citations

1
A Classifier Model Using Fine-Tuned Convolutional Neural Network and Transfer Learning Approaches for Prostate Cancer Detection DOI Creative Commons
Murat Sarıateş, Erdal Özbay

Applied Sciences, Journal Year: 2024, Volume and Issue: 15(1), P. 225 - 225

Published: Dec. 30, 2024

Background: Accurate and reliable classification models play a major role in clinical decision-making processes for prostate cancer (PCa) diagnosis. However, existing methods often demonstrate limited performance, particularly when applied to small datasets binary problems. Objectives: This study aims design fine-tuned deep learning (DL) model capable of classifying PCa MRI images with high accuracy evaluate its performance by comparing it various DL architectures. Methods: In this study, basic convolutional neural network (CNN) was developed subsequently optimized using techniques such as L2 regularization, Tanh activation, dropout, early stopping enhance performance. Additionally, pyramid-type CNN architecture designed simultaneously both fine details broader structures combining low- high-resolution information through feature maps extracted from different layers. approach enabled the learn complex features more effectively. For comparison, enhanced pyramid (FT-EPN) benchmarked against Vgg16, Vgg19, Resnet50, InceptionV3, Densenet121, Xception, which were trained transfer (TL) techniques. It also compared next-generation vision transformer (ViT) MaxViT-v2. Results: The achieved an rate 96.77%, outperforming pre-trained TL like ViT Among models, Vgg19 highest at 92.74%. 93.55%, while MaxViT-v2 95.16%. Conclusions: presents FT-EPN classification, offering reference solution future research. provides significant advantages terms simplicity has been evaluated effective applications.

Language: Английский

Citations

1