Solving the problem of incomplete data in medical diagnosis via interval modeling

• Problem involves making high-quality medical diagnoses when some data are missing.
• Proposed method uses a synergy of many diagnostic models.
• Interval modeling and aggregation significantly improve quality of diagnosis.
• The method has been successfully used in real life diagnosis support system.
• R implementation is available on GitHub for other researchers.

This paper presents an approach to making accurate and high-quality decisions under incomplete information. Our comprehensive approach includes interval modeling of incomplete data, uncertaintification of classical models and aggregation of incomplete results. We conducted a thorough evaluation of our approach using medical data for ovarian tumor diagnosis, where the problem of missing data is commonly encountered. The results confirmed that methods based on interval modeling and aggregation make it possible to reduce the negative impact of lack of data and lead to meaningful and accurate decisions. A diagnostic model developed in this way proved better than classical diagnostic models for ovarian tumor. Additionally, a framework in R that implements our method was created and is available for reproduction of our results. The proposed approach has been incorporated into a real-life diagnosis support system – OvaExpert.

Figure optionsDownload as PowerPoint slide