A team led by Professor Li Hai from the Hefei Institutes of Physical Science, Chinese Academy of Sciences, has recently developed a multimodal deep learning model that can effectively predict the malignancy risk of TI-RADS 4 thyroid nodules with high-risk features. The research results have been published in the journal Computerized Medical Imaging and Graphics, providing a new method for precise diagnosis of thyroid cancer.

The incidence of thyroid cancer has risen rapidly in recent years. Ultrasound examination, as the primary diagnostic tool, is greatly affected by the physician's experience in terms of accuracy. For TI-RADS 4 nodules, traditional diagnostic methods carry a risk of misdiagnosis, which may lead to unnecessary surgery or delayed treatment. The AI model developed by the research team integrates B-mode ultrasound and strain elastography techniques, achieving AUC values of 0.937 on the test set and 0.927 on the external validation set, outperforming single-modality models.

The diagnostic performance of this AI model surpasses that of radiologists. As an auxiliary tool, it can improve the diagnostic accuracy of doctors with varying levels of experience. The heatmaps generated by the model highly align with the regions of interest noted by physicians, confirming its clinical application value. Professor Li Hai stated: "This innovative artificial intelligence model can significantly reduce the risks of misdiagnosis and missed diagnosis, especially for patients with high-risk thyroid cancer."

This study opens a new pathway for intelligent diagnosis of thyroid nodules. The multimodal fusion approach is expected to be extended to other medical imaging fields. In the future, this technology is likely to play an important role in clinical practice, helping doctors make more precise diagnostic decisions.