Related Experiment Video
Updated: Mar 6, 2026
07:00
Identification of OTX1 and OTX2 As Two Possible Molecular Markers for Sinonasal Carcinomas and Olfactory Neuroblastomas
Published on: February 28, 2019
6.4K
MRI-Based Texture Analysis to Differentiate Sinonasal Squamous Cell Carcinoma from Inverted Papilloma.
S Ramkumar1, S Ranjbar2, S Ning1
1From the School of Computing, Informatics, and Decision Systems Engineering (S.Ramkumar, S.N., T.W., J.L.).
AJNR. American Journal of Neuroradiology
|March 4, 2017
Summary
Magnetic resonance imaging texture analysis can differentiate sinonasal squamous cell carcinoma from inverted papilloma. This AI-driven approach shows promise in aiding neuroradiologists in tumor classification.
Area of Science:
- Oncology
- Radiology
- Artificial Intelligence
Background:
- Sinonasal inverted papilloma can harbor squamous cell carcinoma, necessitating accurate tumor differentiation.
- Distinguishing between these two sinonasal tumors is clinically relevant for appropriate patient management.
Purpose of the Study:
- To evaluate the accuracy of MR imaging-based texture analysis in classifying noncoexistent squamous cell carcinoma and inverted papilloma.
- To compare the classification performance of texture analysis with neuroradiologists' interpretations.
Main Methods:
- Texture analysis algorithms were applied to MR imaging data (axial T1, T2, and postcontrast) from resected tumors (>1.5 cm).
- Machine learning models were trained and validated to classify tumor types.
- Two neuroradiologists independently reviewed imaging data (ROI, tumor, entire image) for comparison.
Main Results:
- The best texture analysis model achieved high classification accuracies (training: 90.9%, validation: 84.6%).
- Machine learning accuracy (89.1%) surpassed neuroradiologists' ROI review (56.5%) but was comparable to their review of entire images (87.0%).
Conclusions:
- MR imaging texture analysis demonstrates potential for differentiating sinonasal squamous cell carcinoma from inverted papilloma.
- This technique may offer supplementary diagnostic information to aid neuroradiologists in clinical practice.
