Paper Title
Adaptive Local Threshold With Improved Similarity Based Birchalgorithm For Lymph Node Segmentation In Head and Neck Squamous Cell Carcinoma

Abstract
Head and neck cancer ranks as the seventh most common cancer globally, with squamous cell carcinoma (HNSCC) accounting for 90% of cases. The prognosis of cancer is critically dependent on regional lymph node metastasis. Hence, physicians prescribe Elective Neck Dissection (END) in almost all cases to remove the metastasis and avoid any risk of recurrence. It is often unnecessary, and for some patients, this overtreatment can lead to post-operative morbidity, mainly shoulder dysfunction, and also cause an increased cost of cancer treatment. The selection of patients who will benefit from neck dissection has always been a subject of debate for decades.Existing diagnostic methods are limited by factors such as variability in imaging quality, challenges in feature extraction, inter-observer variability, and high false-positive rates for occult metastasis.Machine learning offers promising solutions by improving diagnosis and treatment planning, requiring precise segmentation of regions of interest. Furthermore, during END, many lymph nodes removed may not contain cancer since the procedure is designed to excise all nodes within a predefined anatomic region. Accurate segmentation algorithms can play a pivotal role in identifying affected areas, thus reducing unnecessary END and improving clinical decision-making.The objective of this study is to design and assess an efficient lymph node segmentation algorithm, which will support clinicians in developing effective treatment strategies. This study proposes an unsupervised algorithm, Adaptive Local Threshold with Improved Similarity-based BIRCH segmentation (ALTIS-BS) and compares its performance with existing algorithms like Fuzzy C-Means, K-Means and conventional BIRCH. Keywords - HNSCC, lymph node segmentation, Machine learning, BIRCH