What question did this study set out to answer?

To develop and validate polytomous risk prediction models classifying ovarian tumors into benign, borderline, primary invasive, or metastatic categories using standard and kernel-based algorithms.

October 20, 2010BMC Medical Research MethodologyOpen Access

Polytomous diagnosis of ovarian tumors as benign, borderline, primary invasive or metastatic: development and validation of standard and kernel-based risk prediction models

View Full Paper

Ask AI

Bookmark

Authors

BCBen Van CalsterRega Institute for Medical Research LVL. ValentinSkåne University Hospital CHCaroline Van HolsbekeAZ Sint-Jan

Discussion

Member takes

Implication

Multi-center validation study reveals effective multi-class ovarian tumor risk prediction in ultrasound patients, highlighting the utility of paired logistic regression.

Key Points

To develop and validate polytomous risk prediction models classifying ovarian tumors into benign, borderline, primary invasive, or metastatic categories using standard and kernel-based algorithms.
Trained diagnostic models using a multicenter development dataset (N=1,066) and tested generalizability on a separate multicenter temporal and external validation dataset (N=1,938).
Compared standard logistic regression against penalized kernel-based algorithms (least squares support vector machines and kernel logistic regression) utilizing true polytomous structures and pairwise coupling.
Conducted variable selection using cross-validated c-indexes and evaluated performance via polytomous and dichotomous c-indexes alongside calibration graphs.
Internal validation achieved polytomous c-indexes between 0.64 and 0.69, with the optimal model yielding dichotomous c-indexes ranging from 0.73 (primary invasive vs metastatic) to 0.96 (borderline vs metastatic).
External and temporal validation demonstrated polytomous c-indexes between 0.57 and 0.64, though discrimination between primary and metastatic invasive tumors decreased to near random levels.
Pairwise coupling of dichotomous standard logistic regression models utilizing 9 to 11 predictors performed comparably to advanced kernel-based algorithms.

Cite This Study

Calster et al. (2010) studied this question.

synapsesocial.com/papers/6a00c07f581c6e761e77d7be https://doi.org/10.1186/1471-2288-10-96

View Full Paper

Ask AI

Bookmark

October 20, 2010BMC Medical Research MethodologyOpen Access

Polytomous diagnosis of ovarian tumors as benign, borderline, primary invasive or metastatic: development and validation of standard and kernel-based risk prediction models

View Full Paper

Ask AI

Bookmark

Authors

BCBen Van CalsterRega Institute for Medical Research LVL. ValentinSkåne University Hospital CHCaroline Van HolsbekeAZ Sint-Jan

Discussion

Member takes

Implication

Multi-center validation study reveals effective multi-class ovarian tumor risk prediction in ultrasound patients, highlighting the utility of paired logistic regression.

Key Points

To develop and validate polytomous risk prediction models classifying ovarian tumors into benign, borderline, primary invasive, or metastatic categories using standard and kernel-based algorithms.
Trained diagnostic models using a multicenter development dataset (N=1,066) and tested generalizability on a separate multicenter temporal and external validation dataset (N=1,938).
Compared standard logistic regression against penalized kernel-based algorithms (least squares support vector machines and kernel logistic regression) utilizing true polytomous structures and pairwise coupling.
Conducted variable selection using cross-validated c-indexes and evaluated performance via polytomous and dichotomous c-indexes alongside calibration graphs.
Internal validation achieved polytomous c-indexes between 0.64 and 0.69, with the optimal model yielding dichotomous c-indexes ranging from 0.73 (primary invasive vs metastatic) to 0.96 (borderline vs metastatic).
External and temporal validation demonstrated polytomous c-indexes between 0.57 and 0.64, though discrimination between primary and metastatic invasive tumors decreased to near random levels.
Pairwise coupling of dichotomous standard logistic regression models utilizing 9 to 11 predictors performed comparably to advanced kernel-based algorithms.

Cite This Study

Calster et al. (2010) studied this question.

synapsesocial.com/papers/6a00c07f581c6e761e77d7be https://doi.org/10.1186/1471-2288-10-96

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

View Full Paper

Ask AI

Bookmark