Purpose This study investigates the efficiency of deep learning models in the automated diagnosis of Hashimoto’s thyroiditis (HT) using real-world ultrasound data from ultrasound examinations by computer-assisted diagnosis (CAD) with artificial intelligence. Methods We retrospectively collected ultrasound images from patients with and without HT from two hospitals in China between September 2008 and February 2018. Images were divided into a training set (80%) and a validation set (20%). We ensembled nine convolutional neural networks (CNNs) as the final model (CAD-HT) for HT classification. The model’s diagnostic performance was validated and compared from two hospital validation sets. We also compared the accuracy of CAD-HT against seniors/junior radiologists. Subgroup analysis of CAD-HT performance in different thyroid hormone levels, such as hyperthyroidism, hypothyroidism, and euthyroidism, was also evaluated. Results 39280 ultrasound images from 21118 patients were included in this study. The accuracy, sensitivity, specificity of the ensemble HT-CAD model were 0.892, 0.890 and 0.895, respectively. HT-CAD performance between two hospitals was not significantly different. The HT-CAD model achieved a higher performance (P < 0.001) when compared to senior radiologists, with a nearly 9% accuracy improvement. HT-CAD had almost similar accuracy (from 0.871 to 0.894) among the three differences of thyroid hormone level subgroups. Conclusion The HT-CAD strategy based on CNN significantly improved the radiologists’ diagnostic accuracy on HT. Our model demonstrates good performance and robustness in different hospitals and for different thyroid hormone levels.

Background: Non-invasive prenatal testing (NIPT) is a commonly employed clinical method to screen for fetal aneuploidy, while the Y chromosome-based NIPT method is regarded as the gold standard for the estimation of fetal fraction (FF) of male fetuses. However, when the fetus has a derivative Y chromosome thereby containing a partial Y chromosome, the Y chromosome-based NIPT method cannot accurately calculate FF. Therefore, alternative methods to precisely calculate FF are required. Methods: Two prenatal cases could not be detected effectively using the Y chromosomebased NIPT method because of low FF. According to the Y chromosome-based method, the FF of the fetuses were 1.730 ± 0.050% (average gestation week: 18 +1) and 2.307 ± 0.191% (average gestation week: 20 +0) for cases 1 and 2, respectively. Using various genetic diagnostic techniques, including the BoBs™ assay, karyotype analysis, improved nucleolus-organizing region (NOR)-banding analysis, Affymetrix CytoScan 750K Array, and fluorescence in situ hybridization (FISH) analysis, we determined the genetic defects of two fetuses with translocations of the SRY locus. Further, we reassessed the FF using FF-QuantSC and X chromosome-based methods. The distribution diagram of reads for chromosome Y was also analyzed. Results: The FF of the fetuses determined by FF-QuantSC were 10.330% (gestation week: 18 +4) in case 1 and 9.470% (gestation week: 21 +4) in case 2, while the FF of the fetuses determined using the X chromosome-based method were 8.889% (gestation week: 18 +4) in case 1 and 2.296% (gestation week: 21 +4) in case 2. Both the distribution diagrams of reads for chromosome Y of the two cases showed the deletion in the long arm of the Y chromosome. Conclusion: For repeatedly low FF samples detected using the Y chromosome-based NIPT method for a long gestational week, we believe that FF-QuantSC and distribution diagrams of reads could be used as a supplement to NIPT, especially for rare cases of sex reversal caused by SRY translocation.