Conţinutul numărului revistei |
Articolul precedent |
Articolul urmator |
![]() |
![]() ![]() |
Ultima descărcare din IBN: 2023-12-21 05:50 |
Căutarea după subiecte similare conform CZU |
004.89:61 (10) |
Искусственный интеллект (314) |
Mедицинские науки (11477) |
![]() . Management of missing values in continuous biomedical data. In: Revista de Ştiinţe ale Sănătăţii din Moldova, 2022, nr. 3 An.1(29), p. 250. ISSN 2345-1467. |
EXPORT metadate: Google Scholar Crossref CERIF DataCite Dublin Core |
Revista de Ştiinţe ale Sănătăţii din Moldova | ||||||
Numărul 3 An.1(29) / 2022 / ISSN 2345-1467 | ||||||
|
||||||
CZU: 004.89:61 | ||||||
Pag. 250-250 | ||||||
|
||||||
![]() |
||||||
Rezumat | ||||||
Background. Primary data are an important pillar of practice and, in particular, of biomedical research. They often contain missing values. For „static” data, several recovery methods are proposed. In the case of data representing continuous biomedical signals, the set of methods is limited. Objective of the study. Presentation of an algorithm for the recovery of continuous biomedical data for later use for machine learning for clinical purposes. Material and Methods. The researched data are publicly available data describing 40,336 patients with sepsis and other pathologies (non-sepsis) provided by the competition „Early Prediction of Sepsis from Clinical Data: the PhysioNet Computing in cardiology Challenge 2019” and contain up to 80.9% of missing data. Results. Using the R programming language, an algorithm was created which, unlike other algorithms (e.g., LOCF – last observation carried forward) considers the dynamics (increase or decrease) of a certain parameter of interest. The data restored using the proposed algorithm are finally used to create a system for early prediction (up to 4 hours before onset) of sepsis, which has a predictive performance of 92% by the area under the ROC curve (AUROC). Conclusion. The proposed algorithm can be used to restore missing values in continuous biomedical data, describing physiological parameters recorded in intensive care units (heart rate, O2 blood stasis, blood pressure, etc.). |
||||||
Cuvinte-cheie missing data recovery, algorithm, machine learning, Artificial Intelligence, restabilirea datelor lipsă, algoritm, învățare automată, inteligenţă artificială |
||||||
|