Article Figures & Data
Figures
- FIG 1.
Flow chart describing preparation of training/validation and test sets. The asterisk indicates that due to its infrequency within the data set (1%) and its similarities with the MR imaging brain without/with contrast protocol, the MR imaging for brain metastasis protocol was combined with the MR imaging brain without/with contrast protocol, reducing the total number of protocols from 9 to 8.
- FIG 2.
Confusion matrix for the LSTM classifier model. Criterion standard labels are on the y-axis. Labels predicted by the model are on the x-axis. Numbers represent the number of examinations in the test set. WO/W indicates MR imaging brain without/with contrast; MS, MR imaging brain for MS; TUMOR, MR imaging tumor; MENINGIOMA, MR imaging brain meningioma follow-up; SEIZURE, MR imaging brain dedicated seizure with contrast; STEALTH, MR imaging stealth/Stryker/mask/presurgical planning; GAMMA, MR imaging for gamma knife, brain lab, and SRS; CN, MR imaging cranial nerves III–VI without/with contrast.
Tables
Training/Validation Set Original Test Set before Protocol Review Final Test Set after Adjudication MR imaging brain without/with contrast 11,543 (68.4%) 1274 (67.9%) 1156 (6.2%) MR imaging brain for MS 2235 (13.2%) 242 (12.9%) 251 (13.4%) MR imaging tumor 963 (5.7%) 113 (6.0%) 149 (7.9%) MR imaging brain meningioma follow-up 653 (3.9%) 81 (4.3%) 117 (6.2%) MR imaging brain dedicated seizure with contrast 519 (3.1%) 65 (3.5%) 65 (3.5%) MR imaging stealth/Stryker/mask/presurgical planning 436 (2.6%) 43 (2.3%) 45 (2.4%) MR imaging for gamma knife, brain lab, and SRS 332 (2.0%) 41 (2.2%) 58 (3.1%) MR imaging cranial nerves III–VI without/with contrast 201 (1.2%) 17 (0.9%) 35 (1.9%) Total 16,882 1876 1876 ↵a Data are No. (%).
Labels Weighted Precision Weighted Recall (Accuracy) Weighted F1 Score Criterion standard labels 0.908 0.905 0.905 Original protocol assignments 0.872 0.841 0.850 ↵a Precision, recall, and F1 score were calculated for each class, and weighted averages of these metrics were computed using class frequencies as weights. The weighted F1 score represents the average of F1 scores weighted by class frequency.
- Table 3:
Contingency table comparing performances of the classifier model and the original protocol assignments on the criterion standard labelsa
Incorrectly Predicted by Original Assignments Correctly Predicted by Original Assignments Incorrectly predicted by model 80 98 Correctly predicted by model 185 1513 ↵a Numbers represent number of examinations in the test set that were incorrectly/correctly classified by the model and incorrectly/correctly classified by the original protocol assignments. The McNemar test of homogeneity statistic is 26.13; P value < .001.
Jump to section
Related Articles
Cited By...
- Evaluating Large Language Model-Generated Brain MRI Protocols: Performance of GPT4o, o3-mini, DeepSeek-R1 and Qwen2.5-72B
- Imaging protocol suggested by large language model depends on language: preliminary experiments using GPT-4
- Real-World Adoption of Artificial Intelligence in Radiology: Opportunities and Barriers