Prediction of breast cancer distant recurrence using natural language processing and knowledge-guided convolutional...

Area of Science:

• Oncology
• Medical Informatics
• Computational Biology

Background:

• Distant recurrence of breast cancer poses significant risks and reduces survival rates.
• Early prediction of distant recurrence is crucial for timely intervention and improved patient outcomes.

Purpose of the Study:

• To design and validate an electronic health record (EHR)-based predictive model for estimating the distant recurrence probability in breast cancer patients.
• To leverage natural language processing (NLP) and machine learning techniques for enhanced breast cancer recurrence prediction.

Main Methods:

• Utilized pathology reports and clinical progress notes from 6,447 breast cancer patients diagnosed between 2001 and 2015.
• Applied NLP tools to map clinical notes to Concept Unique Identifiers (CUI).
• Employed bag-of-words and pre-trained embeddings for feature vectorization, integrating them with structured clinical data for machine learning classifiers and a Knowledge-guided Convolutional Neural Network (K-CNN).

Main Results:

• The developed model achieved an Area Under the Curve (AUC) of 0.888.
• The model demonstrated an F1-score of 0.5, indicating its predictive capability.

Conclusions:

• An automated method for predicting breast cancer distant recurrence using NLP and deep learning has been established.
• The findings suggest potential for improved patient management through early identification of recurrence risk.