Region-of-Interest and Handcrafted vs. Deep Radiomics Feature Comparisons for Survival Outcome Prediction: Application to Lung PET/CT Imaging
Bibliographic record
Abstract
Although many studies have focused on handcrafted radiomics features (RF), a much more indepth study, as follows in this study, is necessary to explore usage of deep RFs extracted from deep learning (DL) algorithms to assess survival analysis. We enrolled 215 lung cancer patients with PET, CT, and clinical data from The Cancer Imaging Archive and the Vancouver General Hospital. This study aims to assess 3 approaches, including comparison of i) PET vs. CT images; ii) different region of interests (ROI) used for deep RF extraction; and iii) deep vs. handcrafted RFs in prediction of overall survival (OS; regression task to predict exact time-to-death and survival probability) by hybrid machine learning systems (HMLS), including Analysis of Variance (ANOVA) and principal component analysis (PCA) linked with regression algorithms (RA). Subsequently, 256 deep RFs were extracted from 3 ROIs on both images by a 3D auto-encoder. The 3 ROIs included i) a segmented lung tumor; ii) a 3D bounding box (36x48x46 millimeter) including the lung tumor area; and iii) entire image. Further, 215 handcrafted RFs were extracted from each segmented tumor on both images through the standardized ViSERA-radiomics package. In OS prediction, i) PET-based HMLSs outperformed CT-based HMLSs, ii) deep RFs extracted from the cropped images outperformed deep RFs extracted from other ROIs, receiving the highest mean absolute error (MAE) of 595±126 [range 36-5596 days] by ANOVA+support vector regression, and iii) deep RFs significantly outperformed handcrafted RFs. In survival probability prediction, i) PET-based HMLSs outperformed CT based HMLSs, ii) PET-based deep RFs extracted from the entire image outperformed deep RFs extracted from other ROIs, having the highest c-index of 0.65±0.05 through PCA+Random Survival Forest, and iii) deep RFs outperformed handcrafted RFs.
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 | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| 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 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".