AI Science Tracker
EN

Learned cervical-cancer recurrence models show modest internal discrimination

Fudan University; Shanghai Key Laboratory of Female Reproductive Endocrine-Related Diseases; Shanghai Changjiang Science and Technology Development · 20 Feb 2021 · medicine

Discovery

Learned cervical-cancer recurrence models show modest internal discrimination

Source check: Source checked. Not independent scientific confirmation.

Finding: In a retrospective four-hospital cervical-cancer analysis, learned models estimated recurrence and survival using recorded clinical and pathological factors. The 1,023-person test set was a random split of the same hospital pool. Site-specific recurrence discrimination was modest, and external clinical benefit was not established.

Technology: Support vector machine and tree-based survival/classification models

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned classifiers and representations evaluate a defined clinical staging or recurrence target; conventional clinicopathologic associations are not separately counted as AI discoveries.
Test and result
The source analyzed 5,112 surgically treated patients, 268 deaths and 343 recurrences, with an 80:20 stratified split, training-only ADASYN and five-fold grid search. For all 22 variables, test recurrence AUC was .794 for SVM versus .785 for logistic regression; site-specific recurrence AUC was .781 versus .776. SVM local/distant test sensitivities were .472/.424 at the represented operating points.
Limitation / still unknown
The source calls the test external, but Methods randomly divide patients pooled from the same four hospitals; this is internal evaluation. Some feature subsets are selected from whole-cohort Cox associations, and containment is not established. Inputs include postoperative pathology/adjuvant treatment, so this is not preoperative prediction or a causal treatment-effect model. Recurrence event counts differ 343 versus 344 across source locations. Binary risk endpoints do not supply a frozen future prediction horizon. Source claims of a universally best model are not supported across all displayed table cells; no clinical surveillance benefit is measured.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC7907920/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 21 Nov 2026. Review scope: Original patient cohort, computational Methods and represented Results actually read.

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 1/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open in the full catalog for filters, charts and related discoveries. Browse medicine discoveries.

Related catalog entries

Links use documented relationships where available, plus shared catalog context. A shared topic or model family does not establish a scientific dependency.