AUTHOR=Wang Ke , Liu Yuehua , Chen Hongxin , Yu Wenjin , Zhou Jiayin , Wang Xiaoying 

TITLE=Fully automating LI-RADS on MRI with deep learning-guided lesion segmentation, feature characterization, and score inference

JOURNAL=Frontiers in Oncology

VOLUME=Volume 13 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2023.1153241

DOI=10.3389/fonc.2023.1153241

ISSN=2234-943X

ABSTRACT=Leveraging deep learning into the radiology com- munity has great potential and practical significance. To explore the potential of fitting deep learning methods into the current LI- RADS system, this paper provides a complete and fully automatic deep learning solution for LI-RADS system and investigates its model performance in liver lesion segmentation and classification. To achieve this, a deep learning study design process is for- mulated, including clinical problem formulation, corresponding deep learning task identification, data acquisition, data prepro- cessing and algorithm validation. An Unet++-based segmentation approach with supervised learning was performed by using 33674 raw images obtained from 113 patients which are collected from 2009 to 2017. A modified VGG16 network with joint training strategy was performed for feature characterization, followed by an inference module to generate the final LI-RADS score. Both liver segmentation and feature characterization models were evaluated respectively and comprehensive statistical analysis was conducted with detailed discussions. Median DICE of liver lesion segmentation was able to achieve 0.879. Based on different thresholds, recall changes within a range of 0.7 to 0.9 and precision always stays high greater than 0.9. Segmentation model performance was also evaluated on patient level and lesion level and the evaluation results of (precision, recall) on patient level was much better around (1,0.9). Lesion classification was evaluated to have an overall accuracy of 76% and most mis- classification cases happen in the neighboring categories which is reasonable since it is naturally difficult to distinguish LI-RADS 4 and LI-RADS 5.