Master’s Thesis Defense of Abbas Abdulkhudhur Mohammed

The Master’s thesis defense of Abbas Abdulkhudhur Mohammed was held in the Department of Biomedical Engineering on Tuesday, September 2, 2025, for his thesis entitled:

“Enhanced Detection of Cervical Spine Fractures using Artificial Intelligence”

The examination committee consisted of:

• Chair: Assoc. Prof. Ikhlas Kazem Hamzah – College of Engineering / University of Technology

• Member: Assoc. Prof. Israa Badr Nasser – College of Engineering / Al-Nahrain University

• Member: Asst. Prof. Alaa A’ed Jabr – College of Engineering / Al-Nahrain University

Supervisors:

• Assoc. Prof. Hadeel Qasim Wadi – College of Engineering / Al-Nahrain University

• Assoc. Prof. Abdul-Latif Ali Asghar Mustafa – College of Medicine / Al-Mustansiriyah University

Scientific Evaluation:

• First Reviewer: Assoc. Prof. Noor Kamal Mohsen – Al-Khwarizmi College of Engineering / University of Baghdad

• Second Reviewer: Assoc. Prof. Shaimaa Safa’ Al-Deen Bahaa Al-Deen – College of Engineering / Al-Nahrain University

Linguistic Review:

• Asst. Prof. Mais Uday Abdulrasool – College of Engineering / Al-Nahrain University

Study Objective:
This research aims to develop a multi-stage framework based on deep learning techniques for the automated detection and localization of cervical spine fractures in computed tomography (CT) images with high accuracy, using a 2.5D approach to integrate contextual information from consecutive slices.

Proposed Framework:

1. Classification using DenseNet121 to detect fracture presence, supported by Grad-CAM for visual interpretability.

2. Segmentation through DenseNet169 to accurately isolate cervical vertebrae (C1–C7).

3. Localization with Faster R-CNN combined with DenseNet169-FPN to draw bounding boxes and pinpoint fracture sites.

Results:
The framework achieved outstanding performance, with:

• Classification accuracy: 98.3%

• Dice similarity coefficient for segmentation: 91.56%

• Overall localization accuracy: 94.98%

Conclusion:
These results confirm that the proposed framework enhances automated diagnosis by improving both the speed and accuracy of cervical spine fracture detection. This reduces the workload on radiologists and contributes to better clinical outcomes and patient care.

The thesis was accepted, having fulfilled all the requirements for the Master’s degree.