A Scoring System for Assessing the Risk of Malignant Partially Cystic Thyroid Nodules Based on Ultrasound Features

Ultrasound-Based Transfer Learning Model to Assist Partially Cystic Thyroid Nodule Diagnosis

PURPOSE

A transfer learning model based on ultrasound was established to predict the malignant probability of partially cystic thyroid nodule (PCTN) preoperatively, providing clinicians with a non-invasive primary screening method.

METHODS

258 PCTNs of 258 patients from January 2020 to January 2024 were analyzed retrospectively. The dataset was randomly divided into a training set and a test set in a ratio of 8:2. Five different pre-trained models were chosen for transfer learning, including EfficientNet, Inception_v3, Mobilenet_v3, SqueezeNet, and VGG19. The area under the curve (AUC), accuracy, sensitivity, and specificity of the training and test cohorts were calculated. Grad-Class Activation Map (Grad-CAM) was drawn to interpret the results visually. All the ultrasound images were reviewed by two radiologists; multivariate logistic analyses explored the independent risk factors for malignant PCTN. The diagnostic effectiveness of transfer learning models and radiologists was compared.

RESULTS

Inception_v3 model achieved the highest AUC of 0.9243 (95% CI: 0.8849-0.9439) in predicting the malignancy of PCTN in the training cohort, with an accuracy of 85.19%, sensitivity of 85.26%, and specificity of 85.00%. The diagnostic efficiency of the Inception_v3 model was better than that obtained by multivariate logistic regression analysis with AUC of 0.8247 (95% CI: 0.7579-0.8915) in the training cohort, with an accuracy of 83.33%, a sensitivity of 68.00%, and a specificity of 71.80%. Red or warm-colored regions in Grad-CAM represented that these features were more important to model decisions, while blue or cool-colored regions represented those features that were less important.

CONCLUSION

Ultrasound-based transfer learning model could predict the malignant probability of PCTN noninvasively before surgery, especially the Inception_v3 model, to assist clinical decision-making.

Comparative analysis of machine learning-based ultrasound radiomics in predicting malignancy of partially cystic thyroid nodules

OBJECTIVE

To investigate the application of machine learning (ML) model-based thyroid ultrasound radiomics in the evaluation of malignancy in partially cystic thyroid nodules (PCTNs).

METHODS

One hundred and ninety-two patients with 197 nodules PCTNs from January 2020 to December 2020 were retrospectively analyzed. Radiomics features were extracted based on hand-crafted features from the ultrasound images, and machine learning methods were used to build a classification model by radiomics features. The least absolute shrinkage and selection operator regression was applied to select the features of nonzero coefficients from radiomics features. The prediction performance of the established model was mainly evaluated by the area under the curve (AUC) and accuracy, sensitivity, and specificity.

RESULTS

Nineteen radiomics features were extracted from the original images for each nodule. Eight ML classifiers were able to differentiate malignancy in PCTNs. The AUC, accuracy, sensitivity, and specificity of k-Nearest Neighbor (KNN) model were 0.909, 82.95%, 83.33%, and 89.90%, respectively, on the test cohort. The comparative result showed statistically equivalent performance for thyroid nodule diagnosis based on image fusion and single image. In addition, the ML-Based ultrasound radiomics system showed a better AUC as compared with ACR TI-RADS model and the ultrasound features model.

CONCLUSION

The novel ultrasonic-based ML model has an important clinical value for predicting malignancy in PCTNs. It can provide clinicians with a preoperative non-invasive primary screening method for PCTN diagnosis to avoid unnecessary medical investment and improve treatment outcomes.

Ultrasound-Based Risk Stratification System for the Assessment of Partially Cystic Thyroid Nodules

OBJECTIVE

To develop and validate a risk stratification system for the prediction of malignancy in Partially Cystic Thyroid Nodules (PCTNs).

METHODS

From January 2020 to December 2021, we retrospectively reviewed the sonography data of PCTNs patients from Hangzhou TCM Hospital, Hangzhou First People's Hospital two medical centers. The independent risk factors of malignant PCTNs were evaluated using univariate and multivariate logistic regression analysis. The nomogram prediction efficiency was assessed using Area Under the Curve (AUC) and calibration curves. To determine the clinic value of the predictive model, Decision Curve Analysis (DCA) was used.

RESULTS

In this retrospective study, a total of 285 patients (301 PCTNs) were enrolled, 242 nodules were benign, and 59 nodules were malignant. Younger age, hypoechoic, irregular margin and microcalcifications were found to be the independent risk factors associated with malignant PCTNs. The AUC, sensitivity, and specificity were 0.860, 77.1%, and 84.7% in the training set and 0.897, 91.7%, and 87.0% in the external validation set, respectively. The total point of nomogram is greater than 161, which shows the best to predict the malignancy of PCTNs.

CONCLUSIONS

Our findings demonstrated that the risk stratification system for the assessment of PTCNs showed good prediction capacities.