Skip to main content
. 2023 Jan 12;15(2):482. doi: 10.3390/cancers15020482

Figure 2.

Figure 2

(A) Classification tasks considered in this study. Task (i) was performed in the entire set of patients included in the study (with WHO grade 2, 3, and 4 gliomas) and in the subset of those with lower-grades (2 and 3) gliomas. Task (ii) was performed only in the subset of patients with lower-grades gliomas. Task (iii) was performed only in the entire set of patients included in the study. (B) Representative conventional and diffusion MR images of a 23-year-old female patient with IDH-mutant and 1p/19q-codeleted oligodendroglioma. The tumour volumes (in red) of each image, obtained after semiautomatic segmentation, were subsequently used as input for the models. (C) Models developed for each task. They differ for the set of images used as input: (a) conventional MRI with tumour laterality (left or right brain hemisphere); (b) diffusion MRI with tumour laterality; (c) conventional and diffusion MRI with tumour laterality. The architecture of the residual network (ResNet10) used to provide the task-specific class prediction was the same. This figure was created with BioRender.com, accessed on 21 December 2022.