PET Inter-Lesion Radiomics Aggregation for Enhanced PRRT Response Prediction in Neuroendocrine Tumors
Bibliographic record
Abstract
Peptide Receptor Radionuclide Therapy (PRRT) using [${ }^{177}$Lu]Lu-DOTA-TATE has significantly improved outcomes for patients with advanced neuroendocrine tumors (NETs), yet predicting therapeutic response remains challenging. This study investigates whether aggregating radiomic features of different lesions extracted from pre-treatment somatostatin receptor PET/CT scans can predict disease progression and time to progression (TTP) in NET patients receiving PRRT. A retrospective analysis was conducted on 81 patients, with segmented lesions sorted based on standardized uptake values ($\text{SUV}_{\text {max }}, \text{SUV}_{\text {mean }}, \text{SUV}_{\text {min }}$) and volume. Radiomic features were extracted from the top 1, 3, and 5 lesions per patient, and two aggregation strategies-stacked and statistical-were applied. Classification models were trained using eight machine learning algorithms incorporating three feature selection methods within a nested cross-validation framework. For TTP prediction, five survival models employing three feature selection methods were used within the same cross-validation scheme. Results showed that stacking features from the top three lesions sorted by$\text{SUV}_{\text {min }}$and input into a K-Nearest Neighbors model provided the highest progression prediction accuracy (AUCC$=0.75$). For TTP, the best performance was achieved by a Random Survival Forest model trained on statistically aggregated features from the top 5 lesions sorted by SUV${ }_{\text {mean }}(\mathrm{C}$-index$=0.68)$. Overall, incorporating radiomic data from multiple lesions using aggregation methods enhanced model performance in both tasks, highlighting the importance of lesion selection and feature aggregation in progression and survival prediction for PRRT-treated NET patients.
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) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| 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 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".