Machine learning-based diagnostics of capsular invasion in thyroid nodules with wide-field second harmonic generation microscopy

• DOI: 10.1016/j.compmedimag.2024.102440
• Publication Date: 2024-10-05T01:47:54Z
• Authors: Yaraslau Padrez; Lena Golubewa; Igor Timoshchenko; Adrian Enache; Lucian G. Eftimie; Radu Hristu; Danielis Rutkauskas
• Keywords: Thyroid Nodules

Papillary thyroid carcinoma (PTC) is one of the most common, well-differentiated carcinomas gland. PTC nodules are often surrounded by a collagen capsule that prevents spread cancer cells. However, as malignant tumor progresses, integrity this protective barrier compromised, and cells invade surroundings. The detection capsular invasion is, therefore, crucial for diagnosis choice treatment development new approaches aimed at increase diagnostic performance great importance. In present study, we exploited wide-field second harmonic generation (SHG) microscopy in combination with texture analysis unsupervised machine learning (ML) to explore possibility quantitative characterization structure designation different areas either intact, disrupted invasion, or apt invasion. Two-step k-means clustering showed capsules all analyzed tissue sections were highly heterogeneous exhibited distinct segments described characteristic ML parameter sets. latter allowed structural interpretation fibers sites overt fragmented curled rarely formed distributed networks. Clustering also distinguished not categorized initial histopathological but could be recognized prospective micro-invasions after additional inspection. features suspicious invasive identified proposed approach can become reliable complement existing methods diagnosing encapsulated PTC, reliability diagnosis, simplify decision making, prevent human-related errors. addition, automated ML-based selection images exclusion non-informative regions greatly accelerate fully integrated into clinical practice.