Academia.eduAcademia.edu

Figure 1 – uploaded by Lequan Yu

See full PDFdownloadDownload figure
Table 2: Results of MICCAI MRBrainS challenge of different methods (DC: %, HD: mm, AVD: %. only top 11 teams are shown here).

Table 2 Results of MICCAI MRBrainS challenge of different methods (DC: %, HD: mm, AVD: %. only top 11 teams are shown here).

Related Figures (11)

Figure 1: Examples of brain images from different image modalities (the first row) and the segmentation annotations by experts (the second row): (a) T2-FLAIR modality of subject2, (b) T1-IR modality of subject2, (c) T1 modality of subject2, (d) T1 modality of subject4 (yellow, green, and red colors represent the WM, GM, and CSF, respectively).
Figure 1: Examples of brain images from different image modalities (the first row) and the segmentation annotations by experts (the second row): (a) T2-FLAIR modality of subject2, (b) T1-IR modality of subject2, (c) T1 modality of subject2, (d) T1 modality of subject4 (yellow, green, and red colors represent the WM, GM, and CSF, respectively).
Figure 2: (a) The architecture of proposed VorResNet for volumetric. image segmentation, consisting of batch normalization layers (BN), rectified linear units (ReLU), and convolu- tional layers N (ConvN) wtih number of channels, filter size and downsampling stride; (b) The illustration of VoxRes module.
Figure 2: (a) The architecture of proposed VorResNet for volumetric. image segmentation, consisting of batch normalization layers (BN), rectified linear units (ReLU), and convolu- tional layers N (ConvN) wtih number of channels, filter size and downsampling stride; (b) The illustration of VoxRes module.
Figure 3: An overview of our proposed framework for integrating auto-context with multi- modality information.
Figure 3: An overview of our proposed framework for integrating auto-context with multi- modality information.
Figure 4: Results of validation data using different modalities: (a)-(c) denote the original T1, TLIR, and T2-FLAIR MR images and (e)-(g) show the corresponding segmentation result using single image modality, respectively; (d) is the ground truth label; (h) is the result using all image modalities without auto-context information (yellow, green, and red colors represent the WM, GM, and CSF, respectively). The blue arrows indicate some incorrect predictions with anatomical structure GM present in (e), WM absent in (f), and GM absent in (g).
Figure 4: Results of validation data using different modalities: (a)-(c) denote the original T1, TLIR, and T2-FLAIR MR images and (e)-(g) show the corresponding segmentation result using single image modality, respectively; (d) is the ground truth label; (h) is the result using all image modalities without auto-context information (yellow, green, and red colors represent the WM, GM, and CSF, respectively). The blue arrows indicate some incorrect predictions with anatomical structure GM present in (e), WM absent in (f), and GM absent in (g).
Table 1: Cross-validation results of MRI brain segmentation using different image modal- ities (DC: %, HD: mm, AVD: %). Figure 5: Results of validation data using all image modalities: (a) original Tl MR im- ages, (b) results of VorResNet, (c) results of Auto-context VorResNet, (d) ground truth labels (yellow, green, and red colors represent the WM, GM, and CSF, respectively). The blue arrows indicate some predictions with anatomical structures mistakenly classified in VorResNet but correct in Auto-contert VorResNet.
Table 1: Cross-validation results of MRI brain segmentation using different image modal- ities (DC: %, HD: mm, AVD: %). Figure 5: Results of validation data using all image modalities: (a) original Tl MR im- ages, (b) results of VorResNet, (c) results of Auto-context VorResNet, (d) ground truth labels (yellow, green, and red colors represent the WM, GM, and CSF, respectively). The blue arrows indicate some predictions with anatomical structures mistakenly classified in VorResNet but correct in Auto-contert VorResNet.
Table 3: Results of MRI brain segmentation using different levels of contextual information (DC: %, HD: mm, AVD: %).
Table 3: Results of MRI brain segmentation using different levels of contextual information (DC: %, HD: mm, AVD: %).
Figure 6: Typical segmentation results of validation data using different levels of con- textual information: (a)-(c) denote the original T1, T1-IR, and T2-FLAIR MR images, respectively; (d) ground truth label; (e)-(h) show the results using Cl, C2, C1-C2, and C1-C4 contextual information, respectively (yellow, green, and red colors represent the WM, GM, and CSF, respectively). The blue arrows indicate some incorrect predictions of anatomical structures with fewer levels contextual information.
Figure 6: Typical segmentation results of validation data using different levels of con- textual information: (a)-(c) denote the original T1, T1-IR, and T2-FLAIR MR images, respectively; (d) ground truth label; (e)-(h) show the results using Cl, C2, C1-C2, and C1-C4 contextual information, respectively (yellow, green, and red colors represent the WM, GM, and CSF, respectively). The blue arrows indicate some incorrect predictions of anatomical structures with fewer levels contextual information.
DC Results of Hippocampus Segmentation Figure 7: Dice results of hippocampus segmentation regarding the number of subjects for training.
DC Results of Hippocampus Segmentation Figure 7: Dice results of hippocampus segmentation regarding the number of subjects for training.
HD Results of Hippocampus Segmentation Figure 8: HD results of hippocampus segmentation regarding the number of subjects for training.
HD Results of Hippocampus Segmentation Figure 8: HD results of hippocampus segmentation regarding the number of subjects for training.
Figure 9: AVD results of hippocampus segmentation regarding the number of subjects for training.
Figure 9: AVD results of hippocampus segmentation regarding the number of subjects for training.
Related topics:
PsychologyComputer ScienceArtificial IntelligenceMedicineMulti ModalityBrain SegmentationPsychology and Cognitive SciencesMedical and Health Sciences
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved