Project ID: P202205130006
Automatic Segmentation of Head and Neck Cancer using Fusion, Attention map detection, and Deep learning Techniques
Deep Learning, Machine Learning, Image Processing
People with the below expertise are able to apply for this project: 1-The individual with enough experience in image processing techniques 2-The individual with enough experience in deep learning segmentation techniques 3- The individual with enough experience in traditional and deep learning fusion techniques 2-The individual with enough experience in attention map detection methods Both programming languages Matlab and Python are acceptable, but Python works better for some people who aim at working on google Colab.
Objectives: Squamous cell carcinoma of the head and neck (HNSCC) is a malignant tumor of the head and neck that originates from the lips, mouth, sinuses, larynx, nasopharynx, and other cancers of the larynx. HNSCC cancer, the sixth fatal cancer, affects over 655,000 people worldwide each year. It is reported that half of these patients resulted in death. In this study, multi-level multi-modality-fusion technique is a promising technique with potential for improved segmentation of head and neck (HN) cancer. We aim to develop automatic segmentation of segment head and neck squamous cell carcinoma (HNSCC) tumors via hybrid systems including deep learning algorithms, attention map algorithms, and fusion techniques. Methods: In this study, we employ 408 patients with head and neck cancer. All patients have PET, CT, its ground truth. Positron emission tomography (PET) images are first registered to computed tomography (CT) images, enhanced, and cropped. First, several fusion techniques are utilized to combine PET and CT information. Moreover, we aim to employ attention map techniques to improve segmentation performance. Thus, attention map detection algorithms are applied to the fused images to emphasis on tumor area. We also aim to utilize multiple deep learning algorithms to automatically segment head and neck squamous cell carcinoma (HNSCC) tumors. Finally, we also apply Ensemble Voting segmented images to enhance segmentation process. 80% of patient data are used for hybrid systems to select the best model based on maximum performance resulting from 5-fold cross-validation. Subsequently, the remaining 20% is used for external testing of the selected model.