AUTHOR=Zhang Yu , Liang Kewei , He Jiaqi , Ma He , Chen Hongyan , Zheng Fei , Zhang Lingling , Wang Xinsheng , Ma Xibo , Chen Xuzhu 

TITLE=Deep Learning With Data Enhancement for the Differentiation of Solitary and Multiple Cerebral Glioblastoma, Lymphoma, and Tumefactive Demyelinating Lesion

JOURNAL=Frontiers in Oncology

VOLUME=Volume 11 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2021.665891

DOI=10.3389/fonc.2021.665891

ISSN=2234-943X

ABSTRACT=Objectives: To explore the differential diagnosis of deep learning based on MRI with data enhancement on brain glioblastoma (GBM), primary central nervous system lymphoma (PCNSL) and tumefactive demyelinating lesion (TDL). 
Materials and Methods: This retrospective study analyzed the MRI data of 261 patients with pathologically diagnosed solitary and multiple cerebral GBM (97), PCNSL (92), and TDL (72). The 3D segmentation model was trained to capture the lesion. Different enhancement data were generated by changing the pixel ratio of the lesion and non-lesion area. The 3D classification network was trained with the enhancement data. The accuracy, sensitivity, specificity and area under curve (AUC) were used to assess the value of different enhancement data on the discrimination performance. The results were compared with those of the neuroradiologists.
Results: The diagnostic performance fluctuated with the change of the ratio of lesion to non-lesion area. When the ratio was 1.5, the diagnostic performance was the best. The AUC of GBM, PCNSL and TDL were 1.00 (95% CI 1.000, 1.000), 0.96 (95% CI 0.923, 1.000) and 0.954(95% CI 0.904, 1.000), respectively. 
Conclusions: The deep learning with data enhancement is useful for the identification of GBM, PCNSL and TDL, and the diagnostic performance outperforms the neuroradiologists.