Region-based nasopharyngeal carcinoma lesion segmentation from MRI using clustering- and classification-based methods...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Clinical diagnosis of nasopharyngeal carcinoma (NPC) requires manual delineation of tumor boundaries on MRI, a process that is time-consuming and expertise-dependent.
• Existing computer-aided segmentation methods often require manual parameter adjustment, posing a challenge for clinicians.
• There is a need for automated and efficient segmentation techniques to support NPC diagnosis.

Purpose of the Study:

• To introduce and evaluate two novel region-based methods for segmenting nasopharyngeal carcinoma (NPC) lesions in magnetic resonance images (MRI).
• To demonstrate the effectiveness of parameter learning in improving NPC segmentation accuracy.
• To provide a more efficient and less burdensome tool for clinicians in NPC diagnosis.

Main Methods:

• Development of two region-based computer-aided segmentation methods incorporating parameter learning.
• Utilized a dataset of 253 MRI slices containing NPC lesions for evaluation.
• Compared the performance of the proposed methods against other existing region-based segmentation techniques.

Main Results:

• The introduced region-based methods with parameter learning demonstrated superior performance in NPC segmentation.
• Experimental results confirmed the advantage of incorporating learning into the segmentation algorithms.
• The methods achieved statistically comparable segmentation performance to other advanced techniques.

Conclusions:

• Parameter learning significantly enhances the performance of region-based methods for NPC segmentation on MRI.
• The developed methods offer a promising solution for automated and accurate tumor delineation, reducing clinical workload.
• These findings support the clinical utility of advanced image analysis techniques in oncology.