A Machine Learning Framework to Predict IQ Decline Post-treatment in Pediatric Brain Cancer with Consideration of Computational Efficiency
Bibliographic record
Abstract
Existing models for predicting intelligence quotient (IQ) for pediatric brain cancer patients after intensity-modulated radiation therapy (RT) are derived from statistical linear and nonlinear analysis. While those models are computationally efficient, they use oversimplified representations of a patient's overall brain features. They neglect other important and complex brain structure information, such as image segmentations of healthy tissue and the target to be treated.\\ Therefore, we develop a machine learning (ML) approach to predict post-RT IQ based on 57 pediatric cases. Our data-model pipeline allows us to train over 600 billion parameters. Such high-density processing pipeline requires a novel parallel computing framework for the training and tuning tasks. Our framework can tractably handle these computational requirements by utilizing 1) an extensive grid search fitting-training scheme on individual and ensemble ML models, 2) automated network morphism that optimizes neural network structure efficiently, and 3) hardware configurations for RAM, CPU and GPU environments. This framework is adaptable to most of the ML-RT applications and particularly useful in both high- and low-dimensional datasets shown in many clinical applications. In predicting high or low post-RT IQ deficit, our framework outperforms baseline models, achieving a test accuracy of 75\%, a mean training accuracy of 83.3\% (±13), recall $\geq75\%$, and precision-recall harmonic mean of 0.75 for pre-Treatment-Planning dataset. We also demonstrate that complex models like ensemble stack model and deep neural nets can be practical under certain conditions.
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 | 0.003 |
| 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 | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 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".