Deep learning model and TI-RADS show higher sensitivity than radiologist assessment

0

[ad_1]

Emerging research suggests that the use of a deep learning algorithm may be beneficial in differentiating between malignant and benign thyroid nodules on ultrasound in children and young adults.

In a retrospective study recently published in the American Journal of Radiology, the researchers looked at data from 139 patients 21 years of age or younger who had a thyroid nodule on ultrasound. Looking at the sensitivity and specificity of differentiating between malignant and benign thyroid nodules, the study authors then compared assessment by an independent radiologist to the use of the reporting system and thyroid imaging data ( TI-RADS) from the American College of Radiology and a previously developed deep learning algorithm.

In contrast to an average sensitivity of 58.3% for independent radiologist impressions, the study authors noted an 87.5% sensitivity for the deep learning algorithm and an 85.1% sensitivity for the use of the TI-RADS classification.

“Given the relatively high sensitivity of the deep learning algorithm, the algorithm may be useful in helping to identify potentially malignant nodules for further evaluation by radiologists in facilities that currently rely solely on the overall impression of radiologists without the use of ACR TI-RADS,” wrote Maciej Mazurowski. , Ph.D., associate professor at Duke University and scientific director of the Duke Center for Artificial Intelligence in Radiology, and colleagues.

However, the researchers noted that the specificity of the deep learning algorithm was significantly lower (36.1%) than using the TI-RADS classification (average of 50.6%) and impression independent of radiologists (average of 79.9%). In light of these results, further training of a deep learning system and further validation in a cohort of children with thyroid nodules are needed before clinical use of a deep learning model in this patient population, Mazurowski and colleagues pointed out.

That said, the study authors note the potential for a deep learning model that could offer a more objective and reproducible approach to evaluating pediatric thyroid nodules in settings where physicians with specific pediatric training do not. are not available.

Mazurowski and colleagues also note that sensitivity in diagnosing thyroid nodules is particularly high in the pediatric population.

“In adults, high specificity in the diagnostic evaluation of thyroid nodules is important given the desire to limit unnecessary biopsies. However, in children, thyroid cancer that is small or mildly aggressive has an extended period of time to grow and/or metastasize. So in children, high sensitivity is a priority,” asserted Mazurowski and colleagues.

Regarding the limitations of the study, the authors acknowledged that the deep learning algorithm was trained with images of adult thyroid nodules. The researchers pointed out that examining radiologists were only asked to give their impression of whether a thyroid nodule was malignant or benign and did not include any possible recommendations for fine-needle aspiration (FNA) in case of a thyroid nodule. diagnostic uncertainty. Mazurowski and his colleagues admitted that the radiologist’s impressions as well as the deep learning algorithm’s assessments were based on two static grayscale images for each nodule. According to the study authors, the high malignancy rate of 40.3% may have been affected by selection bias and potential baseline in a study in a tertiary care setting.

[ad_2]
Source link

Share.

Comments are closed.