مشروع البحث:
Hand fracture diagnosis system using artificial intelligence

صفحة عناصر بسيطة
dc.contributor.advisorد.ابوعجيلة المصري دغمان
dc.date.accessioned2026-06-07T07:55:44Z
dc.date.available2026-06-07T07:55:44Z
dc.descriptionThis thsis focuses on developing an advanced technological tool to bridge the gap between traditional medical diagnosis and modern software engineering. The importance of this research stems from several challenges associated with manual fracture diagnosis, including the complex anatomy of the human hand that complicates the detection of subtle fractures (such as hairline fractures), human factors affecting diagnostic accuracy due to physician fatigue during long work shifts, and the lack of specialized radiologists in some areas, necessitating a clinical decision support system. The idea is to construct a hybrid model that combines unsupervised machine learning and deep neural networks to classify X-ray images into "healthy" or "fractured" with high accuracy. The study followed a meticulous sequential methodology that included data collection, resulting in the creation of a real database from the archives of the radiology department at Yafran General Hospital in Libya. This database comprised images of various fracture cases alongside healthy cases. The data underwent advanced preprocessing to ensure training stability, which included resizing images (224x224 pixels), converting to grayscale, normalizing values, and utilizing data augmentation techniques to avoid overfitting. The proposed model integrates the K-Means algorithm (for data clustering and quality inspection) with Convolutional Neural Networks (CNNs), which are responsible for feature extraction and fracture classification. The development environment leveraged the Python programming language with TensorFlow and Keras libraries for model construction, and the PySide6 library for the development of a graphical user interface (GUI). The results indicat several expected practical and scientific outcomes, including a diagnostic accuracy surpassing 75% in identifying and classifying fractures. The system provides diagnostic results in a timely manner (approximately 3 seconds), thus reducing waiting time in emergency departments. Additionally, a user-friendly interface was developed, allowing physicians to input patient data and receive immediate diagnoses that include a "Confidence Level" for each case, thereby enhancing professional and legal reliability. The system also allows the export of results to Excel files to facilitate data archiving and case tracking. II We emphasize that the integration of machine learning in medical imaging is not merely a technical option, but a necessity for modernizing healthcare services and improving patient treatment outcomes, particularly in areas lacking specialized expertise. Keywords:Hand Fracture Diagnosis ,X-ray Imagery, Deep Learning,Convolutional Neural Networks (CNN),K-Means Clustering,Hybrid,ModelMedical Decision Support System.
dc.description.abstractThis thsis focuses on developing an advanced technological tool to bridge the gap between traditional medical diagnosis and modern software engineering. The importance of this research stems from several challenges associated with manual fracture diagnosis, including the complex anatomy of the human hand that complicates the detection of subtle fractures (such as hairline fractures), human factors affecting diagnostic accuracy due to physician fatigue during long work shifts, and the lack of specialized radiologists in some areas, necessitating a clinical decision support system. The idea is to construct a hybrid model that combines unsupervised machine learning and deep neural networks to classify X-ray images into "healthy" or "fractured" with high accuracy. The study followed a meticulous sequential methodology that included data collection, resulting in the creation of a real database from the archives of the radiology department at Yafran General Hospital in Libya. This database comprised images of various fracture cases alongside healthy cases. The data underwent advanced preprocessing to ensure training stability, which included resizing images (224x224 pixels), converting to grayscale, normalizing values, and utilizing data augmentation techniques to avoid overfitting. The proposed model integrates the K-Means algorithm (for data clustering and quality inspection) with Convolutional Neural Networks (CNNs), which are responsible for feature extraction and fracture classification. The development environment leveraged the Python programming language with TensorFlow and Keras libraries for model construction, and the PySide6 library for the development of a graphical user interface (GUI). The results indicat several expected practical and scientific outcomes, including a diagnostic accuracy surpassing 75% in identifying and classifying fractures. The system provides diagnostic results in a timely manner (approximately 3 seconds), thus reducing waiting time in emergency departments. Additionally, a user-friendly interface was developed, allowing physicians to input patient data and receive immediate diagnoses that include a "Confidence Level" for each case, thereby enhancing professional and legal reliability. The system also allows the export of results to Excel files to facilitate data archiving and case tracking. II We emphasize that the integration of machine learning in medical imaging is not merely a technical option, but a necessity for modernizing healthcare services and improving patient treatment outcomes, particularly in areas lacking specialized expertise. Keywords:Hand Fracture Diagnosis ,X-ray Imagery, Deep Learning,Convolutional Neural Networks (CNN),K-Means Clustering,Hybrid,ModelMedical Decision Support System.
dc.identifier2-9
dc.identifier.urihttps://dspace.academy.edu.ly/handle/123456789/2170
dc.subjectهندسة كهربائية
dc.titleHand fracture diagnosis system using artificial intelligence
dspace.entity.typeProject
project.endDate2025
project.funder.nameالتطبيقية والهندسية
project.investigatorالصادق الطاهر الهمالي
project.startDate2024
الملفات
الحزمة الأصلية
يظهر اﻵن 1 - 1 من 1
لا توجد صورة مصغرة متاحة
اﻻسم:
9-2تشخيص كسور اليد باستخدم خوارزميات التعلم الالي الغير خاضع لا.pdf
الحجم:
2.23 MB
التنسيق:
Adobe Portable Document Format
تحميل
حزمة الترخيص
يظهر اﻵن 1 - 1 من 1
لا توجد صورة مصغرة متاحة
اﻻسم:
license.txt
الحجم:
1.71 KB
التنسيق:
Item-specific license agreed to upon submission
الوصف:
تحميل
مجموعات
مشاريع ورسائل علمية