Optimizing CBCT analysis of the Adenoid region: A deep learning approach
Bibliographic record
Abstract
To develop a deep learning (DL) algorithm to segment the adenoid hypertrophy (AH) area from Cone Beam Computed Tomography (CBCT) scans to aid in the early detection of enlarged adenoids and improve management of AH. This retrospective study utilized CBCT scans, comprising oral radiologist-graded scans for training and validation, and a test dataset diagnosed by an Ear, Nose, and Throat (ENT) specialist using nasoendoscopy (NE), which served as the reference standard for external validation. Manual adenoid area segmentation was performed using 3D Slicer. A DL algorithm, based on convolutional neural networks, was developed to segment the naso- and oropharynx in CBCT images with and without AH. The Dice Similarity Coefficient and Intersection over Union were applied to assess segmentation accuracy. A total of 96 CBCT scans, distributed by AH grading, reflected at least 22,800 DICOM manually segmented files. Evaluator calibration was confirmed within the intraclass correlation coefficient (ICC) 0.90. Data augmentation was applied, maintaining the dataset distribution. Four nnU-Net-based segmentation models were tested: 2D, 3D Full-resolution (3D Fullres), 3D Low-resolution (3D Lowres), and 3D Cascade. The algorithm achieved 0.90 overall accuracy, a 0.90 Dice score, and 0.08 precision on the test dataset for adenoid area segmentation. The trained nnU-Net model demonstrated excellent results in segmenting the AH region, (Dice score: 0.99) achieved for the combination of 3D Cascade and 3D Fullres models. When applied in available imaging, this DL integration with CBCT enhances early AH detection and streamlines referrals for timely treatment by medical teams.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".