A new study published in JAMA Otolaryngology-Head and Neck Surgery investigates whether a non-invasive method of ultrasound imaging, combined with a Google platform machine-learning algorithm, could be used as a rapid and inexpensive first screen for thyroid cancer.
“Currently, ultrasounds can tell us if a nodule looks suspicious, and then the decision is made whether to do a needle biopsy or not,” said Elizabeth Cottrill, MD, an otolaryngologist at Thomas Jefferson University in Philadelphia, and clinical leader of the study. “But, fine-needle biopsies only act as a peephole, they don’t tell us the whole picture. As a result, some biopsies return inconclusive results for whether or not the nodule may be malignant.”
If examining the cells of a needle biopsy proves inconclusive, the sample can be further tested via molecular diagnostics to determine risk of malignancy. However, the standards for when to use molecular testing are still in development, and the test is not yet offered in all practice settings, and is notably absent at many smaller community hospitals.
To improve the predictive power of the ultrasound, Jefferson researchers looked into machine learning models developed by Google. The researchers applied a machine-learning algorithm to ultrasound images of patients’ thyroid nodules to determine whether it could pick out distinguishing patterns.
The researchers trained the algorithm on images from 121 patients who underwent ultrasound-guided fine needle-biopsy with subsequent molecular testing. From 134 total lesions, 43 nodules were classified as high risk and 91 were classified as low risk, based on a panel of genes used in the testing. A preliminary set of images with known risk classifications was used to train the algorithm. From this bank of labeled images, the algorithm used machine-learning technology to pick out patterns associated with high and low risk nodules, and used these patterns to form its own set of internal parameters that could be used to sort future sets of images. The investigators then tested the trained model on a different set of unlabeled images to see how closely it could classify high and low genetic risk nodules.
The researchers found that their algorithm performed with 97% specificity and 90% predictive positive value. The overall accuracy of the algorithm was 77.4%.
Though preliminary, the study suggests that automated machine learning shows promise as an additional diagnostic tool that could improve the efficiency of thyroid cancer diagnoses. Once it becomes more robust, the approach could give doctors and patients more information in order to decide if thyroid lobe removal is necessary.