Relative Importance of Computed Tomography Features Used by Expert Radiologists to Identify Radiologic Patterns in Interstitial Lung Disease
Bibliographic record
Abstract
Abstract Rationale: Clinical practice guidelines define radiologic pattern categories based on integration of multiple individual features; however, it is not clear how important each feature is in the eyes of expert radiologists. Methods: Consecutive patients with fibrotic ILD in the Canadian Registry for Pulmonary Fibrosis were re-evaluated in standardized multidisciplinary discussion. An experienced chest radiologist blinded to clinical data assessed disease distribution and visually quantified the percentage of lung parenchyma affected by honeycombing, reticulation, ground glass, hypoattenuating lung, consolidation, and emphysema. Additional binary features recorded included presence of asymmetric disease, costophrenic angle sparing, 3-density sign, subpleural sparing, dilated esophagus, cysts, and lymphadenopathy. The radiologist then provided a differential diagnosis of patterns with ascribed confidence, mandated to sum to 100%. Patients with confidence >50% for a usual interstitial pneumonia (UIP), fibrotic hypersensitivity pneumonitis (fHP), and non-specific interstitial pneumonia (NSIP) pattern were used in subsequent analyses, compared to cases where no single pattern was >50% (“no confident pattern”). The strength of association of individual radiologic features with radiologist-assigned patterns (UIP, fHP, NSIP) was quantified using a multinomial model, with “no confident pattern” selected as the reference category. Results: 1498 patients were included with patterns of UIP (36%), fHP (17%), NSIP (33%), and “no confident pattern” (14%). Results of the multinomial model are displayed in Figure 1. Increasing honeycombing and reticulation suggested UIP, while ground-glass, hypoattenuating lung, subpleural sparing, and a distribution other than basal and peripheral led away from UIP, including the presence of any central component of disease. Increasing hypoattenuating lung, pure ground-glass, and a 3-density sign suggested fHP, as did a mid-upper predominant distribution, sparing of the extreme costophrenic angle, and the presence of any central component of disease. Axillary lymphadenopathy, a dilated esophagus, and subpleural sparing were strongly associated with NSIP, while increasing honeycombing, reticulation, hypoattenuating lung, and emphysema as well as a non-basal distribution led away from NSIP. The most helpful features (OR≥3 or ≤0.3) were related to distribution or were particularly distinctive findings (e.g. 3-density sign, dilated esophagus, subpleural sparing). Conclusions: Expert radiologists emphasize disease distribution and a few distinctive dichotomous features when determining ILD patterns, likely because these are more reliably identifiable with lower interobserver disagreement compared to continuous variables that lack clear demarcation of clinical significance. These data will help non-experts to assign weights to the building blocks of patterns, increase reproducibility in pattern identification, and improve future clinical practice guidelines in an evidence-based manner.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.002 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 teacher head, 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".