NIRL-08 AUTOMATED LONGITUDINAL TRACKING OF BRAIN METASTASES INTEGRATED IN A USER-ORIENTED SOFTWARE
Bibliographic record
Abstract
Abstract The burden of detection and segmentation of brain metastases (BM) for treatment planning and response assessment has been found to be alleviated by machine learning methods. However, tracking individual lesions over time remains tedious and would benefit from automated assistance for complex cases. We developed a software solution combining an AI-based BM segmentation method with an automated pairing algorithm allowing to track BM across longitudinal scans. The proposed tracking method comprises two steps: identifying lesions in each scan and pairing identified findings across scans. Identification and segmentation of BM is done with our previously published neural network based algorithm. Tracking of BM is achieved by progressively assigning a lesion ID to individual findings. For each series, individual findings are co-registered using image registration to a reference series which defines the initial set of lesions. A matching function then computes a score for each possible finding-lesion pair based on diameter similarity and inter-centroid distance. Findings from highest scoring pairs are sequentially assigned their matched lesion ID, while pairs scoring under a given threshold are assigned a new lesion ID. Series are processed as such until all findings are dispatched. Assignment accuracy was assessed using adjusted rand index (ARI) on a synthetic noisy dataset simulating registration error and misdetections. Findings from 10 to 20 lesions on 5 series of dimension 100 mm per side were randomly generated for 100 synthetic patients each. With a simulated registration error range of 0-2 deg and 0-5mm, the average per patient ARI was 0.94 +/- 0.05, while an error range of 0-4 deg and 0-10mm lowered ARI to 0.87 +/-0.08. Results support the validity of our software solution for automated detection and longitudinal tracking of BM, which can alleviate the burden of follow-ups in the context of stereotactic radiosurgery.
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 | 0.004 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| 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".