Rationale: The aim of this study was to explore possible causal relationships among several variables in the mechanisms of normal and abnormal scar formation using Bayesian networks learning framework. Methods: From January 1994 at the Department of Plastic and Reconstructive Surgery – Burn Center of Traumatological Hospital in Turin clinical histories were constructed for 703 burn patients and 2440 anatomical burn sites by abstraction of details from the clinical notes made during their stay in hospital. A structural and parameter learning algorithm of Bayesian networks was implemented. The Bayesian network was validated on an independent cohort made up of 49 patients and included a total of 162 anatomic burn sites. Results: The main result revealed the relationships between variables showing how the remote variables are linked with the final outcome. Formation of hypertrophic scares is directly related with two variables: time required for the burn to heal and type of burn treatment. However the time to burn healing is related with whole burn area that is in turn in part related with deep burns and age. Finally, age is related with time to first operation, which is associated with the depth of burn. Conclusion: The use of Bayesian networks seemed to facilitate the understanding of the possible relationships among the factors considered.
Bayesian networks for pathological scarring due to burn injuries
BUJA, ALESSANDRA;GREGORI, DARIO;
2009
Abstract
Rationale: The aim of this study was to explore possible causal relationships among several variables in the mechanisms of normal and abnormal scar formation using Bayesian networks learning framework. Methods: From January 1994 at the Department of Plastic and Reconstructive Surgery – Burn Center of Traumatological Hospital in Turin clinical histories were constructed for 703 burn patients and 2440 anatomical burn sites by abstraction of details from the clinical notes made during their stay in hospital. A structural and parameter learning algorithm of Bayesian networks was implemented. The Bayesian network was validated on an independent cohort made up of 49 patients and included a total of 162 anatomic burn sites. Results: The main result revealed the relationships between variables showing how the remote variables are linked with the final outcome. Formation of hypertrophic scares is directly related with two variables: time required for the burn to heal and type of burn treatment. However the time to burn healing is related with whole burn area that is in turn in part related with deep burns and age. Finally, age is related with time to first operation, which is associated with the depth of burn. Conclusion: The use of Bayesian networks seemed to facilitate the understanding of the possible relationships among the factors considered.
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
