|Objective: The increasing use of health self-tracking devices is making the integration of heterogeneous data and shared decision-making more challenging. Computational analysis of lifelog data has been hampered by the lack of semantic and syntactic consistency among lifelog terms and related ontologies. Medical Lifelog Ontology (MELLO) was developed by identifying lifelog concepts and relationships between concepts, and it provides clear definitions by following ontology development methods. MELLO aims to support the classification and semantic mapping of lifelog data from diverse health self-tracking devices.
Methods: MELLO was developed using the General Formal Ontology method with a manual iterative process comprising five steps: (1) defining the scope of lifelog data, (2) identifying lifelog concepts, (3) assigning relationships among MELLO concepts, (4) developing MELLO properties (e.g., synonyms, preferred terms, and definitions) for each MELLO concept, and (5) evaluating representative layers of the ontology content. An evaluation was performed by classifying 11 devices into 3 classes by subjects, and performing pairwise comparisons of lifelog terms among 5 devices in each class as measured using the Jaccard similarity index.
Results: MELLO represents a comprehensive knowledge base of 1,998 lifelog concepts, with 4,996 synonyms for 1,211 (61%) concepts and 1,395 definitions for 926 (46%) concepts. The Web-based MELLO Browser and MELLO Mapper provide convenient access and annotating non-standard proprietary terms with MELLO (http://mello.snubi.org/).
Conclusions: MELLO is the first ontology for representing health-related lifelog data with rich contents including definitions, synonyms, and semantic relationships. MELLO fills the semantic gaps among heterogeneous lifelog terms that are generated by diverse health self-tracking devices. The unified representation of lifelog terms facilitated by MELLO can help describe lifestyle of an individual and environmental factors, which can be included with user-generated data for clinical research and thereby enhance data integration and sharing.
2. MELLO Statistics
|MELLO Concept Type||First-level Term Names||No. of MELLO Perferred Terms||No. of MELLO Terms||Maximum Depth|
|MELLO Primary Terms||Behavior||115||335||7|
|Total No. of MELLO Primary Terms||659||2,270||9|
|MELLO Secondary Terms||Body region||73||296||4|
|Lifelog apps & devices||50||47||2|
|Total No. of MELLO Secondary Terms||1,339||2,726||6|
3. Information Source Map
3.1 Composition of each source data in MELLO
3.2 Venn diagrams indicating how MELLO concepts were derived from various sources
|(A) MELLO primary terms of MELLO (n=659) are primarily derived from SNOMED-CT (n=517), HVV (n=216), and proprietary lifelog terms (n=325). Only 0.16% of SNOMED-CT terms are related to lifelog terms. |
(B) Lifelog qualifier terms of MELLO (n=1,339) are primarily extracted from UMLS (n=1,243), including SNOMED-CT terms, HVV (n=217), and proprietary lifelog terms (n=1,266). Only 0.06% of UMLS terms are related to lifelog terms.