Related Experiment Video
Updated: Sep 11, 2025
Evaluation of Biomarkers in Glioma by Immunohistochemistry on Paraffin-Embedded 3D Glioma Neurosphere Cultures
Published on: January 9, 2019
Clinical evaluation of two glioblastoma delineation methods based on neural networks
Anders Traberg Hansen1, Johannes Thestrup Askglæde1, Jesper Folsted Kallehauge2,3
1Department of Oncology, Aarhus University Hospital, Aarhus, Denmark.
Background And Purpose:
Precise gross tumour volume definition is essential for radiotherapy. Neural networks may improve tumour delineation and reduce manual workload. However, clinical evaluation is crucial for understanding their precision and limitations.
Materials And Methods:
Two neural network-based models were evaluated for glioblastoma delineation in 70 clinical cases: one developed by Cercare Medical Inc (CMN) and the publicly available Raidionics model. Delineations were compared using Hausdorff 95% (HD95) distance, Dice similarity coefficient (DSC) and the prevalence of false-positive and false-negative volumes. Additionally, interobserver variability between clinicians and the dosimetric consequences of differences in delineation were assessed.
Results:
The Raidionics model achieved a mean HD95 of 5.61 mm, with a 5th and 95th percentile range of 2.13-14.8 mm, and a mean DSC of 0.80 [0.62, 0.92]. The CMN model achieved a mean HD95 of 4.24 mm [2.05, 10.2] and mean DSC of 0.83 [0.65, 0.93]. For both metrics the Wilcoxon rank test showed a significant difference (p < 0.002). Both models produced small false-positive volumes, averaging less than 10 % of the true volume. The false-negative volumes averaged around 20 % of the true tumour volume for both models. The HD95 and DSC of interobserver variability were found to be 2.91 mm and 0.89 respectively.
Conclusion:
The CMN performed significantly better than the Raidionics model. Both models demonstrated a low occurrence of false-positive delineations and acceptable robustness in preserving dose coverage. However, their performance remained inferior to clinical experts. Further model development is recommended before potential clinical implementation.
Insights
Neural networks can aid radiotherapy tumor delineation, but clinical evaluation is key. The Cercare Medical Inc (CMN) model showed superior performance over the Raidionics model, though both lagged behind expert clinicians.
Area of Science:
- Radiotherapy
- Medical Imaging
- Artificial Intelligence
Background:
- Accurate gross tumor volume (GTV) definition is critical for effective radiotherapy planning.
- Neural networks offer potential for automated tumor delineation, aiming to enhance precision and reduce manual segmentation workload.
- Clinical validation is essential to ascertain the accuracy and limitations of these AI-driven tools.
Purpose of the Study:
- To evaluate and compare the performance of two neural network-based models for glioblastoma delineation.
- To assess the precision of automated delineations against manual segmentations by clinicians.
- To analyze the dosimetric impact of discrepancies in tumor delineation.
Main Methods:
- Two AI models, Cercare Medical Inc (CMN) and Raidionics, were applied to 70 clinical glioblastoma cases.
- Delineation accuracy was quantified using Hausdorff 95% (HD95) distance and Dice Similarity Coefficient (DSC).
- Interobserver variability among clinicians and dosimetric consequences were also assessed.
Main Results:
- The CMN model achieved superior results with a mean HD95 of 4.24 mm and DSC of 0.83, compared to Raidionics (mean HD95: 5.61 mm, DSC: 0.80).
- Both models exhibited low false-positive rates (<10% of true volume) but higher false-negative rates (~20% of true volume).
- Interobserver variability showed HD95 of 2.91 mm and DSC of 0.89, indicating expert performance was superior to AI models.
Conclusions:
- The CMN model demonstrated significantly better performance than the Raidionics model in glioblastoma delineation.
- While both AI models showed acceptable robustness for dose coverage, their accuracy did not match that of clinical experts.
- Further development and validation are recommended prior to widespread clinical adoption of these neural network models.
More Related Videos
05:45Quantitative Immunohistochemistry of the Cellular Microenvironment in Patient Glioblastoma Resections
Published on: July 31, 2017
09:53Quantifying the Brain Metastatic Tumor Micro-Environment using an Organ-On-A Chip 3D Model, Machine Learning, and Confocal Tomography
Published on: August 16, 2020