Patient AUPENN-GBM-00359 · HIGH RISK

01

Patient overview

AGE

79 years

SEX

Female

KPS

Not recorded

DIAGNOSIS

GBM, IDH-wildtype, WHO Grade IV

MGMT STATUS

Methylated

EXTENT OF RESECTION

Gross total resection

IDH STATUS

IDH-wildtype

COHORT

UPenn-GBM

02

Survival prediction

MEDIAN8.9 mo95% CI6.9–10.9

03

Top features driving this prediction

VIEW

MRI · multi-plane viewer with tumor overlay

SLICE — · — · —

04

AI reasoning trace

PI

Principal investigator

Methodology recommendation

▾

Patient presents an outwardly favorable clinical profile (methylated MGMT, gross total resection) but advanced age (79) and radiomic phenotype place the case in the top risk decile. T1c rim-to-core ratio and FLAIR heterogeneity are both >1.5 SD above the cohort mean, consistent with aggressive growth despite the favorable molecular signature. Recommended model weighting: emphasize penalized Cox and gradient boosting models which capture clinical-radiomic interactions in the older cohort. Decrease weight on deep survival models which underpredict risk for older patients whose clinical fields look favorable on the surface.

RS

Researcher

Ensemble execution + weights

▾

Executed 8-model Super Learner ensemble. Survival-curve monotonicity satisfied across all base learners. Concordance index on training fold: 0.671. Integrated Brier Score: 0.187.

CoxNet	0.31
GBSA	0.22
RSF	0.18
Clinical-only	0.14
XGB-AFT	0.08
DeepSurv	0.04
DeepHit	0.02
DSM	0.01

CR

Critic

Verification verdict

▾

PASS · STATISTICAL CHECK

Train-test C-index gap 0.008 (threshold 0.05). No evidence of overfitting.

PASS · CLINICAL PLAUSIBILITY

Methylated MGMT correctly weighted protective (HR 0.54); age effect positive (1.04/yr); radiomic risk score appropriately dominates in this case. All directions align with established GBM prognostic literature.

PASS · REPRODUCIBILITY

Feature selection (mRMR k=30) and ComBat harmonization applied per-fold. No data leakage detected. TRIPOD+AI reporting elements present.

05

Clinical summary

This 79-year-old patient with IDH-wildtype, MGMT-methylated GBM achieved gross total resection but the model places the case in the top risk decile based on advanced age and aggressive radiomic phenotype. Predicted median survival 8.9 months (95% CI shown above) is shorter than the cohort median of 14.6 months despite a clinically favorable surgical and molecular profile. Recommend early discussion of treatment intensity, supportive-care goals, and clinical-trial eligibility given the within-stratum risk identified by the radiomic features.

Patient BUPENN-GBM-00534 · MEDIUM RISK

01

Patient overview

AGE

62 years

SEX

Female

KPS

Not recorded

DIAGNOSIS

GBM, IDH-wildtype, WHO Grade IV

MGMT STATUS

Methylated

EXTENT OF RESECTION

Gross total resection

IDH STATUS

IDH-wildtype

COHORT

UPenn-GBM

02

Survival prediction

MEDIAN13.8 mo95% CI11.3–16.1

03

Top features driving this prediction

VIEW

MRI · multi-plane viewer with tumor overlay

SLICE — · — · —

04

AI reasoning trace

PI

Principal investigator

Methodology recommendation

▾

Patient presents a balanced-risk profile: age 62 near the cohort median, methylated MGMT (favorable), gross total resection achieved. Radiomic features show mildly elevated peritumoral FLAIR heterogeneity but otherwise fall within the cohort interquartile range. Recommended model weighting: distribute evenly across the ensemble; this patient sits near the training distribution centroid where all base learners perform comparably.

RS

Researcher

Ensemble execution + weights

▾

Executed 8-model Super Learner ensemble. NNLS weights show balanced contribution across clinical and ML models. Concordance index on training fold: 0.668. Integrated Brier Score: 0.171.

CoxNet	0.22
GBSA	0.18
RSF	0.17
Clinical-only	0.16
XGB-AFT	0.12
DeepSurv	0.08
DeepHit	0.05
DSM	0.02

CR

Critic

Verification verdict

▾

PASS · STATISTICAL CHECK

Train-test C-index gap 0.012 (threshold 0.05). Model well-calibrated for this risk stratum.

PASS · CLINICAL PLAUSIBILITY

Methylated MGMT protective (HR 0.54). GTR strongly protective (HR 0.19). Radiomic harm signal modest and consistent with FLAIR-only literature.

PASS · REPRODUCIBILITY

Same preprocessing pipeline as cohort. No data leakage detected. TRIPOD+AI reporting elements present.

05

Clinical summary

This 62-year-old patient with IDH-wildtype, MGMT-methylated GBM achieved gross total resection. The ensemble model predicts a median survival of 13.8 months, placing the case at the cohort median. Mild radiomic FLAIR heterogeneity warrants close imaging surveillance. Recommend standard Stupp protocol with consideration of TTFields adjuvant therapy and routine follow-up imaging.

Patient CUPENN-GBM-00380 · LOW RISK

01

Patient overview

AGE

56 years

SEX

Male

KPS

Not recorded

DIAGNOSIS

GBM, IDH-wildtype, WHO Grade IV

MGMT STATUS

Methylated

EXTENT OF RESECTION

Gross total resection

IDH STATUS

IDH-wildtype

COHORT

UPenn-GBM

02

Survival prediction

MEDIAN21.8 mo95% CI17.8–27.5

03

Top features driving this prediction

VIEW

MRI · multi-plane viewer with tumor overlay

SLICE — · — · —

04

AI reasoning trace

PI

Principal investigator

Methodology recommendation

▾

Patient presents a favorable clinical profile: age 56 (below cohort median), methylated MGMT, gross total resection. Radiomic features show below-median enhancing tumor volume and unremarkable FLAIR heterogeneity. Recommended model weighting: emphasize ensemble members that handle the favorable tail well; the clinical-only baseline and CoxNet historically perform best in this stratum. Reduce weight on tree-based models which can saturate at low risk.

RS

Researcher

Ensemble execution + weights

▾

Executed 8-model Super Learner ensemble. Weighting biased toward clinical-only baseline and penalized Cox, which dominate in low-risk strata. Concordance index on training fold: 0.683. Integrated Brier Score: 0.142.

CoxNet	0.34
Clinical-only	0.27
GBSA	0.14
RSF	0.10
XGB-AFT	0.07
DeepSurv	0.05
DeepHit	0.02
DSM	0.01

CR

Critic

Verification verdict

▾

PASS · STATISTICAL CHECK

Train-test C-index gap 0.005 (threshold 0.05). Strong generalization in favorable stratum.

PASS · CLINICAL PLAUSIBILITY

All four protective factors (young age, methylated MGMT, GTR, favorable radiomic phenotype) carry HR < 1 with literature-consistent magnitudes.

PASS · REPRODUCIBILITY

Same per-fold preprocessing and harmonization. No data leakage detected. TRIPOD+AI reporting elements present.

05

Clinical summary

This 56-year-old patient with IDH-wildtype, MGMT-methylated GBM has multiple converging favorable factors: relatively young age, gross total resection, and a low-volume enhancing tumor on imaging. The ensemble model predicts a median survival of 21.8 months, placing the case in the bottom risk decile of this cohort. Recommend full-intensity Stupp protocol with consideration of clinical-trial enrollment given favorable biology.