A Novel Neutrosophic Method for Automatic Seed Point Selection in Thyroid Nodule Images

Area of Science:

• Medical Imaging
• Endocrinology
• Computer Vision

Background:

• Thyroid nodules are a common endocrine issue requiring accurate detection for improved patient survival.
• Manual and semi-automatic segmentation methods for thyroid nodule detection suffer from accuracy limitations due to factors like sonographer experience and operational latency.
• Homogeneous lesion regions in ultrasound images present challenges for traditional segmentation techniques.

Purpose of the Study:

• To develop and evaluate a novel, automated procedure for seed point selection in thyroid nodule images.
• To enhance the accuracy and efficiency of thyroid nodule segmentation in ultrasound images.
• To improve the diagnostic value of ultrasound imaging for thyroid nodule malignancy assessment.

Main Methods:

• A three-component system was proposed: neutrosophic image enhancement for speckle reduction, an automatic seed selection algorithm based on Higher Order Spectra Entropies (HOSE) from Radon Transform (RT), and region growing segmentation with a constant threshold.
• The automatic seed selection leverages the principle that HOSE values from RT at different angles fall within a specific entropy range.
• Speckle noise was reduced using neutrosophic image enhancement techniques.

Main Results:

• The proposed automated segmentation method achieved high performance metrics: True Positive (TP) value of 96.44 ± 3.01%, False Positive (FP) value of 3.55 ± 1.45%, Dice Coefficient (DC) of 92.24 ± 6.47%, Similarity Index (SI) of 80.57 ± 1.06%, and Hausdroff Distance (HD) of 0.42 ± 0.24 pixels.
• Comparison with other methods demonstrated the superior accuracy and reliability of the proposed system.
• The method effectively addresses the challenges posed by homogeneous lesion regions in ultrasound images.

Conclusions:

• The developed automated segmentation system offers a significant advancement in thyroid nodule detection and characterization.
• This technique can serve as a valuable tool for endocrinologists, potentially improving the accuracy of malignancy diagnosis in thyroid nodules.
• The proposed method enhances the utility of ultrasound imaging in the clinical management of thyroid nodules.