Noninvasive Glioma Grading with Deep Learning: A Pilot Study

Abstract

Gliomas are the most common neuroepithelial brain tumors, different by various biological tissue types and prognosis. They could be graded with four levels according to the 2007 WHO classification. The emergence of non-invasive histological and molecular diagnostics for nervous system neoplasms can revolutionize the efficacy and safety of medical care and radically reduce healthcare costs. Our pilot study aimed to evaluate the diagnostic accuracy of deep learning (DL) in subtyping gliomas by WHO grades (I–IV) based on preoperative magnetic resonance imaging (MRI) from Burdenko Neurosurgery Center’s database. A total of 707 MRI studies was included. A “3D classification” approach predicting tumor type for the entire patient’s MRI data showed the best result (accuracy = 83%, ROC AUC = 0.95), consistent with that of other authors who used different methodologies. Our preliminary results proved the separability of MR T1 axial images with contrast enhancement by WHO grade using DL.

Performance of deep learning algorithms to distinguish high-grade glioma from low-grade glioma: A systematic review and meta-analysis

“… 31 and Decuyper et al., 32 while others defined Ⅲ and Ⅳ(24/54) grades as HGG, such as Danilov et al. 11 and Li et al. 16 Here, our study found that Ⅳ represented HGG in classification was similar to Ⅲ+Ⅳ(SE: 94% vs. 93%; SP:93% vs. 94%; AUC 0.98 vs. 0.98).…”

“…with 1773 images (only 58% SE and 78% SP). 11 Thus, the heterogeneity was still high in the > 130 sample size group, whereas it was decreased in the ≤ 130 group (≤ 130: I 2 = 74.08% in SE and 80.42% in SP; > 130: I 2 = 99.05% in SE and 97.73% in SP). Our study implied that data quality varied enormously in the glioma classification area, which inevitably hindered us from drawing the conclusion of DL.…”

“…(2022) 10 Newly diagnosed with a glioma and when preoperative pre- and post-contrast T1w, T2w, and T2w-FLAIR scans were available The absence of one (or more) of the required scans (T1, post-contrast T1, T2w, T2w-FLAIR) 1748 NR(NR; NR) Danilov et al. (2022) 11 NR NR 707 NR(NR; NR) Chen et al. (2022) 12 Patients diagnosed with glioma after case diagnostic screening; patients with complete imaging and follow-up data; and patients with complete follow-up records Patients with other malignant tumors at the same time; patients with other serious underlying diseases or with dysfunction of important organs such as the heart, lung, liver, and kidney; those who died of diseases or accidents other than glioma; and those who suffered from claustrophobia.…”

“…(2022) 10 Brain tumor Histopathology Random split-sample validation Yes No Danilov et al. (2022) 11 Brain tumor Histopathology Random split-sample validation No No Chen et al. (2022) 12 Brain tumor Histopathology NR No No Tripathi et al.…”

“…(2022), 10 MRI NR No CNN No Retrospective study, data rom 4 in-house datasets and 5 publicly available datasets 6032/NR/960 1508/NR/240 NR Yes Danilov et al. (2022), 11 MRI NR No DenseNet, Resnest200e No Retrospective study, data from N.N. Burdenko Neurosurgery Center, Russia 15957/1773/NR 636/71/NR 2009–2018 No Chen et al.…”