TY - GEN
T1 - Bayesian Network to Support Diagnosis of Rare Diseases in Chile
AU - Araya, David
AU - Marquez, Javier
AU - Nakousi, Nicole
AU - Taramasco, Carla
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - Contrary to popular belief, rare genetic diseases affect a significant portion of the global population, with a prevalence ranging from 3.5% to 8 %. These conditions are particularly prevalent among children. In Chile, the interval between the onset of symptoms and the diagnosis of these diseases can extend between six and eight years, resulting in significant emotional and economic costs for affected families. The primary reason for this delay is the dearth of knowledge about these diseases among neuropediatricians. To address this issue, we propose the implementation of a clinical decision support system (CDSS) called Diagen-AI which infers the condition or disease based on information from the child's phenotype (symptoms and signs). In addition to aiding in the diagnosis, the system offers suggestions regarding potential tests and facilitates the integration of Chilean physicians' expertise with statistical data on clinical conditions documented in Orphanet and insights from scientific literature, a novel approach for this type of solution. Diagen-AI operates by employing a Bayesian network to estimate the posteriori probability associated with the likelihood of a given condition based on observed symptoms. Testing recommendations are derived from the estimation of the impact of incorporating a new test as supplementary evidence in the prediction of conditions. The validation of Diagen-AI was conducted through the generation of synthetic data. Preliminary results were successful, for example, if the algorithm is informed with 50% of the symptoms, a correct diagnosis is achieved in 80% of the cases. With respect to the recommendation of tests, it is verified that on average 3 visits to the doctor (with 3 tests per visit) are required to achieve a correct diagnosis in 80% of the cases. We believe that Diagen-AI will be a valuable tool to shorten diagnostic periods, reducing the suffering and uncertainty of affected families by generating synthetic data.
AB - Contrary to popular belief, rare genetic diseases affect a significant portion of the global population, with a prevalence ranging from 3.5% to 8 %. These conditions are particularly prevalent among children. In Chile, the interval between the onset of symptoms and the diagnosis of these diseases can extend between six and eight years, resulting in significant emotional and economic costs for affected families. The primary reason for this delay is the dearth of knowledge about these diseases among neuropediatricians. To address this issue, we propose the implementation of a clinical decision support system (CDSS) called Diagen-AI which infers the condition or disease based on information from the child's phenotype (symptoms and signs). In addition to aiding in the diagnosis, the system offers suggestions regarding potential tests and facilitates the integration of Chilean physicians' expertise with statistical data on clinical conditions documented in Orphanet and insights from scientific literature, a novel approach for this type of solution. Diagen-AI operates by employing a Bayesian network to estimate the posteriori probability associated with the likelihood of a given condition based on observed symptoms. Testing recommendations are derived from the estimation of the impact of incorporating a new test as supplementary evidence in the prediction of conditions. The validation of Diagen-AI was conducted through the generation of synthetic data. Preliminary results were successful, for example, if the algorithm is informed with 50% of the symptoms, a correct diagnosis is achieved in 80% of the cases. With respect to the recommendation of tests, it is verified that on average 3 visits to the doctor (with 3 tests per visit) are required to achieve a correct diagnosis in 80% of the cases. We believe that Diagen-AI will be a valuable tool to shorten diagnostic periods, reducing the suffering and uncertainty of affected families by generating synthetic data.
KW - Bayesian network
KW - CDSS
KW - rare disease
UR - http://www.scopus.com/inward/record.url?scp=85207848380&partnerID=8YFLogxK
U2 - 10.1109/CLEI64178.2024.10700061
DO - 10.1109/CLEI64178.2024.10700061
M3 - Conference contribution
AN - SCOPUS:85207848380
T3 - Proceedings - 2024 50th Latin American Computing Conference, CLEI 2024
BT - Proceedings - 2024 50th Latin American Computing Conference, CLEI 2024
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 50th Latin American Computing Conference, CLEI 2024
Y2 - 12 August 2024 through 16 August 2024
ER -