Prediction of 10-year Overall Survival in Patients with Operable Cervical Cancer using a Probabilistic Neural Network

Area of Science:

• Oncology
• Machine Learning
• Medical Informatics

Background:

• Predicting long-term cancer survival is crucial for guiding treatment decisions.
• Cervical cancer survival prediction can be enhanced using advanced computational models.
• Surgical treatment outcomes for cervical cancer require reliable prognostic tools.

Purpose of the Study:

• To evaluate the predictive capability of a probabilistic neural network (PNN) for 10-year overall survival in cervical cancer patients.
• To assess the PNN model's performance against established statistical methods like logistic regression and decision trees.
• To determine the utility of PNN in clinical decision-making for cervical cancer treatment.

Main Methods:

• Utilized a dataset of 102 cervical cancer patients (FIGO stage IA2-IIB) treated with radical hysterectomy.
• Input variables included demographic, tumor-related, and perioperative parameters.
• Employed DTREG software for computer simulations and PNN model development, comparing results with logistic regression and decision trees.

Main Results:

• The PNN model demonstrated high predictive accuracy with a sensitivity of 0.949 and specificity of 0.679.
• Achieved a low error rate of 12.5% and a high area under the receiver operating characteristic curve (0.809).
• PNN performance surpassed that of logistic regression and single decision tree models.

Conclusions:

• The probabilistic neural network model reliably predicts 10-year overall survival in women with operable cervical cancer.
• PNN serves as a valuable tool for enhancing decision-making in cervical cancer treatment planning.
• This study highlights the potential of machine learning in improving cancer prognostication and patient management.