eduzhai > Applied Sciences > Engineering >

Learning Tumor Growth via Follow-Up Volume Prediction for Lung Nodules

  • Save

... pages left unread,continue reading

Document pages: 10 pages

Abstract: Follow-up serves an important role in the management of pulmonary nodules forlung cancer. Imaging diagnostic guidelines with expert consensus have been madeto help radiologists make clinical decision for each patient. However, tumorgrowth is such a complicated process that it is difficult to stratify high-risknodules from low-risk ones based on morphologic characteristics. On the otherhand, recent deep learning studies using convolutional neural networks (CNNs)to predict the malignancy score of nodules, only provides clinicians withblack-box predictions. To this end, we propose a unified framework, namedNodule Follow-Up Prediction Network (NoFoNet), which predicts the growth ofpulmonary nodules with high-quality visual appearances and accuratequantitative results, given any time interval from baseline observations. It isachieved by predicting future displacement field of each voxel with a WarpNet.A TextureNet is further developed to refine textural details of WarpNetoutputs. We also introduce techniques including Temporal Encoding Module andWarp Segmentation Loss to encourage time-aware and shape-aware representationlearning. We build an in-house follow-up dataset from two medical centers tovalidate the effectiveness of the proposed method. NoFoNet significantlyoutperforms direct prediction by a U-Net in terms of visual quality; moreimportantly, it demonstrates accurate differentiating performance between high-and low-risk nodules. Our promising results suggest the potentials in computeraided intervention for lung nodule management.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×