While some authors argue that ontological realism should be relaxed in some aspects of ontology modelling
] due to its apparent over-complexity, the creation of ad-hoc new primitive classes has unforeseeable downstream consequences. One of the main benefits of the realist approach is to allow modelling convergence despite domain-specific and application-specific perspective differences by using scientific results and the interdisciplinary bridging perspective of philosophical ontology as a methodology to arrive at more precise and unambiguous ontological structure as a substitute for unexamined natural language assertions such as form the strategy behind terminological resources such as MeSH. Representations of the world according to the consensual scientific discourse guarantee reliable and robust representation artefacts
]. Our description of process attributes maintains this principle, but at the same time provides sufficient expressivity to meet the domain requirements.
When analysed more deeply, process attributes as used in domain terminologies reveal themselves to be a loosely related set of descriptions, reifications and analogies used to communicate some characteristics of events in natural language. We provide in Figure
a comparison between SNOMED, PATO, VSO and the present approach.
Comparison between different process qualities representations.
The most basic characteristic of the heart cycle is the fact that it is a cycle and can be described according to its frequency and variation of sequential periods. A frequency of a cycle refers to its cardinality within some given time which can itself be fully expressed using primitive ontological constructs. Likewise, the duration of periods are real entities and require no special construct. Time intervals are occurrents, just as processes, and therefore not qualities. Durations of time intervals are numeric values.
Despite the underlying compatibility with primitive constructs, a full definition of attributes in our use case was possible only in some cases due to limitations in the underlying logical description language. The right degree of expressivity of a logical language has been subject to very long dispute. However, due to its balance between expressivity, decidability, and most importantly, due to its being an official W3C recommendation, the Web Ontology Language (OWL-DL) is the de facto
standard for representing ontologies across many domains including that of biomedicine. OWL is based on description logics, which are carefully selected decidable subsets of first order logic. Our main difficulty in defining some attributes (such as acceleration of heart rate) results from the lack of generalised support for arithmetical relations on individuals, such as greater than
and less than
, which are needed in our acceleration examples between instances of ordered cycles. Therefore, we could not assert that cycle n
immediately precedes cycle n+1
and has a greater duration than cycle n+1
While not completely representable in OWL, we argue that the FOL definition we provided for some process attributes still contributes to restricting possible interpretations of the class. We can, for instance, distinguish between an strictly accelerating heart beating (every cycle is longer than the one it precedes), a constant accelerating cycle (every cycle is longer than the one it precedes, being the difference between any 2 sequential cycles duration always the same) and an accelerated heart beating, or palpitation (as defined in 3.c.i). The distinctions were shown to be quite adequate for representing the general meaning of common expressions and categories for cardiac arrhythmias, when tested against the criteria. It is important to emphasise, however, that natural language use of these words is much more relaxed and context-dependent, which may require local adaptations according to application needs. Also, many distinctions require the proper identification and classification of types of cycle, which is a rich subject area of its own and out of scope for the current work. Finally, the FOL definition cannot be used for reasoning purposes, since consistency checking is done by OWL reasoners (classification and consistency checking). While our representation could be extended to more expressive logics, this would raise additional concerns regarding decidability that are not in the scope of this paper.
A special case is the description of rhythm patterns like the bigeminal3 rhythm. While at first classified under patterned period variation – since its rhythm is easily recognisable, with a short cycle followed by a long cycle followed by a short cycle, and so on – it can also be described according to the origin of the electrical impulse leading to heart contraction (supra-ventricular and ventricular).
Application to non-cyclical use cases
It is important to highlight that the mentioned design patterns apply exclusively to cyclic processes. However, our approach was developed for generically representing attributes aside from cyclic process. In several cases, we have to decompose a complex process in order to understand what an attribute intends to describe. For example, a pain process can be understood as the summation of nociceptor stimuli. Here, however, not the duration of the action potential matters but the frequency with which action potentials are produced by a group of nociceptors. Allowing the exact description of the process does not mean that such a precise measurement is possible in clinical practice. As discussed in
], separation between the fact and information about the fact can be used to properly describe this situation (using the OWL ‘only’ operator). Also, many pain-related entities common to clinical practice are epistemological entities, which must be carefully evaluated for suitability in realist ontologies
Pregnancy is another highly complex process, due to its mutually coordinated structural and functional changes in (at least) two organisms. The pregnancy process, focusing on the mother’s organism is commonly dissected by fiat into three trimesters, whereas the development of the offspring is split into embryogenesis and foetal development. The sub-process that terminates the pregnancy is the delivery, which again, can be split into a series of processes, such as the sequence of configuration of the baby’s head and body within the birth channel, and the progress of the mother’s labour. The variants of the pregnancy process are manifold in terms of
of the whole process, or process parts, such as labour of repetitive phases such as uterine contractions in relation to the intermittent latency phases
number of contractions
· extra process parts
surgical interventions such as episiotomy or caesarean section
complications of pregnancy such as eclampsia or diabetes
· missing process parts
failure of descent of the foetal head
The pregnancy is also characterised by its participants (mother, offspring), their related body parts and qualities, such as number of offspring, their size, missing or supernumerary parts etc.
Due to the myriad of determinants of a pregnancy process, a classification into “normal” and “abnormal” cannot be reduced to hard criteria. Apart from some extreme situations (e.g. foetal death, miscarriage), the boundary between the normal and the abnormal is fuzzy, as is common in medicine. We argue that the correct description of participants and sub-processes allow proper comparison of different abnormalities, without the arbitrary creation of different terms. Our approach promotes the precise description of each occurrent and participant of the pregnancy process, in order to maintain modelling coherence and accurate representation. For instance, premature labour could be defined according to the time span between each contraction cycle or the cardinality of contraction within a given time span, and the occurrence of these contractions within the time interval spanning from conception to the 37th week after conception. Therefore, it is clear what makes normal labour and premature labour pregnancies similar and what is the distinction between the normal and pathological process parts.
As pointed out by the heart cycle example, the logical language (in this case OWL-DL) imposes limits on what can be adequately represented therein. The proper evaluation of this limitation in ontological representation and reasoning remains to be evaluated. However, it is important to point out that ontological analysis here proposed is independent of particular representations, and is coherent with the philosophical view put forward in BFO foundational papers. Particularly, this approach is coherent with the view that “processes do not change, because processes are changes” put forward by Smith
]. It is also compatible with BFO 2.0, which introduces Process profile
as a special sort of processual parts
]. In this paper, we do not propose a different interpretation, but have rather outlined a complementary approach than a proper ontological definition of complex processes. Therefore, instead of determining profiles according to an ad hoc
structural dimension of a process, process attributes require a precise definition in terms of the kinds of participants, participant qualities and sub-processes that characterize the (attributed) process.