{"componentChunkName":"component---src-templates-blog-post-js","path":"/blog/a-novel-ultrasound-image-diagnostic-method-for-thyroid-nodules/","result":{"data":{"site":{"siteMetadata":{"title":"No Frills News"}},"contentfulNfnPost":{"postTitle":"A novel ultrasound image diagnostic method for thyroid nodules","slug":"a-novel-ultrasound-image-diagnostic-method-for-thyroid-nodules","createdLocal":"2023-01-31 14:31:03.150532","publishDate":"None","feedName":"Image Recognition","sourceUrl":{"sourceUrl":"https://www.nature.com/articles/s41598-023-28932-2"},"postSummary":{"childMarkdownRemark":{"html":"<p>The following figure shows the gradient-weighted class activation maps of the two groups of nodule ultrasound classification.\nThe small nodule classification performance of the model still needs to be guaranteed, so this paper adopts a dual-branch structure to complete the nodule classification task.\nIn this paper, a ResNet-based dual-branch classification network is used, and the nodular area image is cropped by using the positioning result.\nResNet101 shows moderate specificity but the highest accuracy and sensitivity for uncut ultrasound images of nodules.\nWhen training the classification network, the nodular region dataset generated by the uncut dataset and the positioning network is used.</p>"}}}},"pageContext":{"slug":"a-novel-ultrasound-image-diagnostic-method-for-thyroid-nodules"}}}