A Comprehensive Evaluation of CNN and Transformer Models for Automated Bone Fracture Detection

Muş Alparslan Üniversitesi Fen Bilimleri Dergisi, Journal Year: 2024, Volume and Issue: unknown

Published: Aug. 15, 2024

The most significant component of the skeletal and muscular system, whose function is vital to human existence, are bones. Breaking a bone might occur from specific hit or violent rearward movement. In this study, fracture detection was performed using convolutional neural network (CNN) based models, Faster R-CNN RetinaNet, as well transformer-based model, DETR (Detection Transformer). A detailed investigation conducted different backbone networks for each model. This study's primary contributions methodical assessment performance variations between CNN transformer designs. Models trained on an open-source dataset consisting 5145 images were tested 750 test images. According results, RetinaNet/ResNet101 model exhibited superior with 0.901 mAP50 ratio compared other models. obtained results show promising outcomes that models could be utilized in computer-aided diagnosis (CAD) systems.

Language: Английский

A Review of Datasets, Optimization Strategies, and Learning Algorithms for Analyzing Alzheimer’s Dementia Detection

Neuropsychiatric Disease and Treatment, Journal Year: 2024, Volume and Issue: Volume 20, P. 2203 - 2225

Published: Nov. 1, 2024

Alzheimer's Dementia (AD) is a progressive neurological disorder that affects memory and cognitive function, necessitating early detection for its effective management. This poses significant challenge to global public health. The accurate of dementia crucial several reasons. First, timely facilitates intervention planning treatment. Second, precise diagnostic methods are essential distinguishing from other disorders medical conditions may present with similar symptoms. Continuous analysis improvements in have contributed advancements research. It helps identify new biomarkers, refine existing tools, foster the development innovative technologies, ultimately leading more efficient approaches dementia. paper presents critical multimodal imaging datasets, learning algorithms, optimisation techniques utilised context detection. focus on understanding challenges employing diverse modalities, such as MRI (Magnetic Resonance Imaging), PET (Positron Emission Tomography), EEG (ElectroEncephaloGram). study evaluated various machine deep models, transfer techniques, generative adversarial networks multi-modality data In addition, examination encompassing algorithms hyperparameter tuning strategies processing analysing images presented this discern their influence model performance generalisation. Thorough enhancement fundamental addressing healthcare posed by dementia, facilitating interventions, improving accuracy, advancing research neurodegenerative diseases.

Language: Английский