Image-driven classification of functioning and nonfunctioning pituitary adenoma by deep convolutional neural networks

• DOI: 10.1016/j.csbj.2021.05.023
• Publication Date: 2021-05-14T16:10:27Z
• Authors: Hongyu Li; Qi Zhao; Yihua Zhang; Ke Sai; Lunshan Xu; Yonggao Mou; Yubin Xie; Jian Ren; Xiaobing Jiang
• Keywords: 03 medical and health sciences; 0302 clinical medicine; 0202 electrical engineering, electronic engineering, information engineering; Deep learning; 02 engineering and technology; Pituitary adenomas; TP248.13-248.65; MRI; Biotechnology; Research Article

The secreting function of pituitary adenomas (PAs) plays a critical role in making the treatment strategies. However, Magnetic Resonance Imaging (MRI) analysis for is labor intensive and highly variable among radiologists. In this work, by applying convolutional neural network (CNN), we built segmentation classification model to help distinguish functioning from non-functioning subtypes with 3D MRI images 185 patients PAs (two centers). Specifically, adopts concept transfer learning uses pre-trained extract deep features conventional images. As result, both models obtained high performance two internal validation datasets an external testing dataset (for model: Dice score = 0.8188, 0.8091 0.8093 respectively; AUROC 0.8063, 0.7881 0.8478, respectively). addition, considers attention mechanism better interpretation. Taken together, work provides first learning-based tumor region PAs, which enables early diagnosis subtyping