Objective:

To improve the classification of thyroid disorders using a modified DenseNet-201 model applied to Tc-99m scintigraphy images, enhancing diagnostic accuracy and patient outcomes.

Key Findings:

• The study classified seven thyroid disorders, including cold nodule, hot nodule, multi-nodular goiter, nodular goiter, thyroiditis, toxic diffuse goiter, and normal, achieving an accuracy improvement of X% (insert specific metric).
• The modified DenseNet-201 model improved classification accuracy from low-resolution scintigraphy images, demonstrating significant advancements in diagnostic precision.
• AI-driven approaches, particularly CNNs, enhance diagnostic precision in thyroid nodule evaluation, leading to better clinical decision-making.

Interpretation:

The integration of deep learning techniques in analyzing scintigraphy images can lead to more accurate diagnoses of thyroid disorders, potentially improving patient outcomes through timely and appropriate treatment.

Limitations:

• Lower image resolution in scintigraphy may still affect diagnostic accuracy; future work should explore image enhancement techniques.
• Manual analysis by nuclear physicians can introduce human error; automated systems could mitigate this risk.
• Limited availability of nuclear medicine physicians may impact the implementation of the proposed model; training programs could help address this shortage.

Conclusion:

The study demonstrates the potential of deep learning to enhance the classification of thyroid disorders, addressing limitations of traditional diagnostic methods and suggesting implications for improved clinical practice.