Abstract:
In recent years, image analysis has become crucial across various fields, from medical
diagnostics to autonomous vehicles. At the core of image analysis lies segmentation, a critical
step that influences the performance of subsequent processes such as object detection and
recognition. This study focuses on heart tumor segmentation in CT images using deep learning,
particularly convolutional neural networks (CNNs). We utilized the U-Net architecture,
renowned for its precision in medical image segmentation, and introduced a modified version
called Law-Net, which incorporates Laws filters.Our work is divided into four chapters: an
introduction to medical imaging and segmentation techniques, a detailed overview of U-Net
architecture, an exploration of Laws filters and the modified U-Net, and an experimental analysis
of our model's performance. By applying various segmentation evaluation metrics, we
demonstrated the effectiveness of our approach in accurately detecting and segmenting heart
tumors.The results indicate that our model, leveraging widely available data and optimized
parameters, achieved high performance across all evaluation metrics. This confirms the potential
of deep learning in aiding objective medical diagnoses and underscores its capacity to enhance
the accuracy and efficiency of tumor detection in CT images. Our deep learning model holds
promise for significantly improving the diagnostic process for medical professionals.