In order to gather clinical data, a part of the BOUNCE project is user experiments. This way data can be gathered and studied in order to improve understanding and the capacity to predict the resilience of women. The goal is to achieve more personalized interventions and eventually contribute to clinical outcomes and patient well-being.

Validating BOUNCE AI models

Within the BOUNCE project, a clinical validation of the developed Artificial Intelligence (AI) models, incorporated in the BOUNCE Clinical Decision Support System (CDSS), was conducted using a pilot study comprising 60 breast cancer patients.

The aim of this validation was to demonstrate valid clinical evidence of the usability of the provided CDSS towards assessing whether the AI models improves clinicians’ performance to predict patients’ resilience during the treatment process.

QoL prediction

The experiment setup was piloted with an AI model focusing on breast cancer patients’ quality of life (QoL) prediction after six months from the start of treatment. Six clinicians from HUS and three clinicians from CHAMP participated in the validation using BOUNCE CDSS and predicted QoL for 60 breast cancer patients with and without the aid of the implemented AI models.

Predicting patients’ resilience

Our preliminary results showed that clinicians’ performance to evaluate the patients’ QoL was higher with the aid of machine learning predictions than without the aid according to the Area under the Curve (AUC) performance metric. AUC of clinicians was 0.777 with the aid and 0.755 without the aid. When the model’s prediction was correct, the average accuracy (ACC) of the clinicians was .788 with the aid and .717 without the aid.

Our next steps involve experiments from longer follow-up data from multiple stakeholders including HUS, CHAMP and IEO.
The receiver operating characteristic (ROC) curves for machine learning model and clinicians with/without the aid of machine learning prediction

Share This

Share this post with your friends!