Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Sep 21:43:100677.
doi: 10.1016/j.ctro.2023.100677. eCollection 2023 Nov.

Comparing supervised and semi-supervised machine learning approaches in NTCP modeling to predict complications in head and neck cancer patients

I Spiero  1 , E Schuit  1 , O B Wijers  2 , F J P Hoebers  3 , J A Langendijk  4 , A M Leeuwenberg  1
Affiliations

Comparing supervised and semi-supervised machine learning approaches in NTCP modeling to predict complications in head and neck cancer patients

I Spiero et al. Clin Transl Radiat Oncol. .

Abstract

Background and purpose: Head and neck cancer (HNC) patients treated with radiotherapy often suffer from radiation-induced toxicities. Normal Tissue Complication Probability (NTCP) modeling can be used to determine the probability to develop these toxicities based on patient, tumor, treatment and dose characteristics. Since the currently used NTCP models are developed using supervised methods that discard unlabeled patient data, we assessed whether the addition of unlabeled patient data by using semi-supervised modeling would gain predictive performance.

Materials and methods: The semi-supervised method of self-training was compared to supervised regression methods with and without prior multiple imputation by chained equation (MICE). The models were developed for the most common toxicity outcomes in HNC patients, xerostomia (dry mouth) and dysphagia (difficulty swallowing), measured at six months after treatment, in a development cohort of 750 HNC patients. The models were externally validated in a validation cohort of 395 HNC patients. Model performance was assessed by discrimination and calibration.

Results: MICE and self-training did not improve performance in terms of discrimination or calibration at external validation compared to current regression models. In addition, the relative performance of the different models did not change upon a decrease in the amount of (labeled) data available for model development. Models using ridge regression outperformed the logistic models for the dysphagia outcome.

Conclusion: Since there was no apparent gain in the addition of unlabeled patient data by using the semi-supervised method of self-training or MICE, the supervised regression models would still be preferred in current NTCP modeling for HNC patients.

Keywords: Head and neck cancer; NTCP modeling; Radiation-induced toxicity; Semi-supervised learning.

PubMed Disclaimer

Conflict of interest statement

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: J.A. Langendijk reports a relationship with Dutch Cancer Society: funding grants. J.A. Langendijk reports his department has research contracts with IBA, RaySearch, Siemens, Elekta, Leoni, and Mirada. J.A. Langendijk reports a relationship with Global Scientific Advisory Board of IBA, RayCare International Advisory Board of RaySearch that includes: board membership, consulting or advisory, and speaking and lecture fees. J.A. Langendijk reports a relationship with Netherlands Society for Radiation Oncology that includes: board membership. Funding AL is funded by the European Union’s Horizon 2020 research and innovation programme under grant agreement N° 825162. This dissemination reflects only the author's view and the Commission is not responsible for any use that may be made of the information it contains.

Figures

Fig. 1
Fig. 1
External validation of the models for xerostomia grade ≥ 2 for different amounts of labeled data. The amount of unlabeled data is fixed at 40 observations. (a) The AUCs, (b) the calibration intercepts, and (c) the calibration slopes.
Fig. 2
Fig. 2
External validation of the models for dysphagia grade ≥ 2 for different amounts of labeled data. The amount of unlabeled data is fixed at 40 observations. (a) The AUCs, (b) the calibration intercepts, and (c) the calibration slopes.
See this image and copyright information in PMC

References

    1. Langendijk J.A., Doornaert P., Verdonck-de Leeuw I.M., Leemans C.R., Aaronson N.K., Slotman B.J. Impact of late treatment-related toxicity on quality of life among patients with head and neck cancer treated with radiotherapy. J Clin Oncol. 2008;26(22):3770–3776. - PubMed
    1. Kierkels R.G.J., Korevaar E.W., Steenbakkers R.J.H.M., Janssen T., van’t Veld A.A., Langendijk J.A., et al. Direct use of multivariable normal tissue complication probability models in treatment plan optimisation for individualised head and neck cancer radiotherapy produces clinically acceptable treatment plans. Radiother Oncol. 2014;112(3):430–436. - PubMed
    1. Langendijk J.A., Lambin P., De Ruysscher D., Widder J., Bos M., Verheij M. Selection of patients for radiotherapy with protons aiming at reduction of side effects: the model-based approach. Radiother Oncol. 2013;107(3):267–273. - PubMed
    1. Protonentherapie LP, Landelijk H-h. Platform Radiotherapy, Landelijk Indicatie Protocol Protonentherapie, versie 2.2 (LIPPv2.2). 2019.
    1. Van Engelen J.E., Hoos H.H. A survey on semi-supervised learning. Mach Learn. 2020;109(2):373–440.

LinkOut - more resources