Related Experiment Video
Updated: Mar 14, 2026
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
Prospective validation of the ultrasound based TIRADS (Thyroid Imaging Reporting And Data System) classification:
Eleonora Horvath1, Claudio F Silva2, Sergio Majlis3
1Radiology Department, Clinica Alemana, Facultad de Medicina Clinica Alemana, Universidad del Desarrollo, Santiago, Chile. eleonora.horvath@gmail.com.
Objective:
To assess performance of TIRADS classification on a prospective surgical cohort, demonstrating its clinical usefulness.
Methods:
Between June 2009 and October 2012, patients assessed with pre-operative ultrasound (US) were included in this IRB-approved study. Nodules were categorised according to our previously described TIRADS classification. Final pathological diagnosis was obtained from the thyroidectomy specimen. Sensitivity, specificity, positive/negative predictive values and likelihood ratios were calculated.
Results:
The study included 210 patients with 502 nodules (average: 2.39 (±1.64) nodules/patient). Median size was 7 mm (3-60 mm). Malignancy was 0 % (0/116) in TIRADS 2, 1.79 % (1/56) in TIRADS 3, 76.13 % (185/243) in TIRADS 4 [subgroups: TIRADS 4A 5.88 % (1/17), TIRADS 4B 62.82 % (49/78), TIRADS 4C 91.22 % (135/148)], and 98.85 % (86/87) in TIRADS 5. With a cut-off point at TIRADS 4-5 to perform FNAB, we obtained: sensitivity 99.6 % (95 % CI: 98.9-100.0), specificity 74.35 % (95 % CI: 68.7-80.0), PPV 82.1 % (95 % CI: 78.0-86.3), NPV 99.4 % (95 % CI: 98.3-100.0), PLR 3.9 (95 % CI: 3.6-4.2) and an NLR 0.005 (95 % CI: 0.003-0.04) for malignancy.
Conclusion:
US-based TIRADS classification allows selection of nodules requiring FNAB and recognition of those with a low malignancy risk.
Key Points:
• TIRADS classification allows accurate selection of thyroid nodules requiring biopsy (TIRADS 4-5). • The recognition of benign/possibly benign patterns can avoid unnecessary procedures. • This classification and its sonographic patterns are validated using surgical specimens.
