Radiofrequency ultrasound and an artificial intelligence (AI) model can be used to effectively predict the malignancy of thyroid nodules, as well as stratify their risk, according to a study set for publication in the November issue of Ultrasonics.
The combination of radiofrequency ultrasound with an artificial neural network (ANN) could also avoid operator dependency issues and help prevent unnecessary thyroid biopsies, according to a group led by Dr. Chunrui Liu of the Affiliated Hospital of Nanjing University Medical School in China.
"The proposed method has no operator dependency; all of the analyses are performed by computer," the team noted (Ultrasonics, November 2019, Vol. 99, pp. 1-9). "Preliminary results indicated that the performance of ANN combined with radiofrequency ultrasound signals is better than that combined with conventional ultrasound images."
Common but not often malignant
Thyroid nodules are common, but only 8% to 16% are actually malignant, according to the researchers. Many ultrasound techniques are used to evaluate nodule malignancy, including strain elastography, acoustic radiation force impulse imaging, and contrast-enhanced ultrasound, but these methods' efficacy remains unclear, the group wrote.
That's where radiofrequency ultrasound comes in. The technique elicits more clinical information than conventional ultrasound by extracting radiofrequency signals from tissues. But how it performs with thyroid nodules has not been studied, Liu and colleagues noted.
"Preliminary studies of radiofrequency ultrasound have been promising, and the method has been shown to have broader prospective applications in identifying prostate and breast cancers and grading fatty liver," they wrote. "To date, few studies on radiofrequency ultrasound's thyroid cancer detection performance have been reported."
The researchers developed their method to predict suspicious thyroid nodules by first gathering radiofrequency data and then creating radiofrequency ultrasound images using Matlab software (MathWorks). After a radiologist outlined regions of interest on the images, textural features were then analyzed using the gray-level co-occurrence matrix (GLCM) algorithm and principal component analysis. The resulting characteristic values from the textural analysis were subsequently used to train the ANN.
The study included 131 pathologically proven thyroid nodules, of which 59 were benign and 72 were malignant. The nodules were randomly divided into training, validation, and testing cohorts. To test their hypothesis that radiofrequency ultrasound could provide more tissue characteristic information than conventional ultrasound, the researchers also performed the same texture and ANN analyses on the B-mode ultrasound images.
The ANN algorithm performed better with radiofrequency ultrasound than it did on conventional ultrasound in all categories except specificity, the group found.
| ANN performance for predicting thyroid nodule malignancy | ||
| Performance measure | ANN on conventional ultrasound images | ANN on radiofrequency ultrasound images |
| Sensitivity | 94.4% | 100% |
| Specificity | 93.2% | 91.5% |
| Accuracy | 93.9% | 96.2% |
| AUC* | 0.917 | 0.945 |
The group also used the ANN with radiofrequency ultrasound to characterize new malignancy risk groups for categories 3 (probably benign), 4 (suspicious), and 5 (probably malignant) thyroid nodules as established by the American College of Radiology's Thyroid Imaging Reporting and Data System (TI-RADS). The new categories better distinguished malignant nodules compared with TI-RADS.
| Parameter | Category 3 | Category 4 | Category 5 |
| No. of samples | 42 | 16 | 73 |
| No. of malignant samples | 0 | 3 | 69 |
| Risk of malignancy | |||
| ANN plus radiofrequency ultrasound | 0 | 18.8% | 94.5% |
| TI-RADS | 0 | 55.1% | 88.2% |
"The new categories allow for a selection of suspicious nodules to be submitted to fine-needle aspiration, thereby avoiding unnecessary thyroid biopsies," the group wrote.
More research to come
More research needs to be done to establish the benefits of using an ANN and radiofrequency ultrasound, according to Liu and colleagues.
"Of course, although these preliminary results suggested [the use of the ANN and radiofrequency ultrasound] could help sonographers to identify risky thyroid nodules and reduce the number of unnecessary thyroid biopsies, more data will be collected and analyzed in our future study to further confirm the feasibility and accuracy of the proposed method," they concluded.
![Examples of ultrasound findings and techniques. (A) Images in a 39-year-old male patient with a mass in the left thigh. The mass is heterogeneous on the B-mode US image (compared with the patient in D) and showed increased microvascularity (superb microvascular imaging [SMI]) and shear-wave elastography (SWE) values. Undifferentiated pleomorphic sarcoma was diagnosed at biopsy (with pleomorphic rhabdomyosarcoma in surgical specimen). (B) Images in an 18-year-old male patient with a mass in the left leg. The mass is hypoechoic on the B-mode image, with no other findings suggestive of malignancy. The lesion is in contact with the cortex of the tibia, which is slightly irregular. CT revealed a doubtful anteromedial tibial erosion. The microvascular study demonstrated high vascularization, suggestive of malignancy. Periosteal Ewing sarcoma was diagnosed with both histologic and immunohistochemical confirmation. (C) Images in a 69-year-old female patient with a lump growing on the outside of the left leg. Multiple SWE examinations were performed (please note the high values obtained in the measurements, whereas the color map highlights the stiffness relative to adjacent tissues). SMI showed areas of increased vascularization to target for sampling. Undifferentiated spindle cell sarcoma was diagnosed at biopsy, with residual leiomyosarcoma in the surgical specimen after neoadjuvant therapy. (D) Images in a 56-year-old female patient with a mass in the right thigh. The mass is heterogeneous at both B-mode ultrasound (similar to patient A) and MRI (coronal T2-weighted spectral attenuated inversion recovery [SPAIR]; T1-weighted pre-contrast and postcontrast imaging), which even shows uptake after the administration of paramagnetic contrast material, which is traditionally suggestive of malignancy. Low values at SMI and elastography are suggestive of benignity. Spindle cell lipoma was diagnosed at biopsy, with atypical spindle cell lipomatous tumor in the surgical specimen.](https://img.auntminnie.com/mindful/smg/workspaces/default/uploads/2026/08/images-radiol250278fig2.APCFLSvX6p.jpg?auto=format%2Ccompress&dpr=2&fit=crop&h=167&q=70&w=250)
