Improving Thyroid Testing with AI and Image Analysis
Author Information
Author(s): Fairooz Towfeeq, McNamee Sara E., Finlay Dewar, Ng Kok Yew, McLaughlin James
Primary Institution: Ulster University
Hypothesis
Can textural features and hybrid AI models enhance the sensitivity of lateral flow assays for thyroid diagnosis?
Conclusion
The study achieved over 95% accuracy in classifying thyroid-stimulating hormone levels using advanced image analysis techniques.
Supporting Evidence
- The proposed method achieved accuracy levels exceeding 95%.
- Machine learning classifiers demonstrated strong specificity (>99%) in identifying true negatives.
- AHMO-GLCM features provided a significant advantage over traditional intensity-based approaches.
Takeaway
This study shows that using special computer techniques can help doctors test for thyroid problems more accurately and quickly.
Methodology
The study used a novel Averaged Horizontal Multiple Offsets Gray-Level Co-occurrence Matrix to analyze images from lateral flow assays.
Potential Biases
Potential overfitting due to the limited diversity of the dataset.
Limitations
The high dimensionality of features may lead to longer processing times, and the study focused solely on TSH analysis.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website