WinAlgos: software for algologists - Pathos

The need to use information technology to collect patients' clinical data and to store them in convenient paperless archives, where they can be quickly re-examined during subsequent visits and compared for research purposes, led one of the authors (Guido Orlandini) to design an Algological Clinical File in 2001, which was then implemented and named WinAlgos by an IT expert (engineer Mauro Montagni) in a very detailed version which (also translated into English) was much appreciated and used not only in Italy, but at the same time was little used because it was considered too complex at a time when familiarity with computers was still rather limited.1

Twenty years later, WinAlgos (simplified) has been transformed from a simple clinical file into an application that not only guides the algological examination, but also serves as a tool for the guided formulation of the pathogenetic and syndromic diagnosis of pain.

The study of the topography of pain involves choosing from a figure of human body the somatic region(s) involved, together with the affected side (Figure 3) and assessing the superficial or deep character and type of topographical distribution.

The study of pain characters involves the detection of character of intensity pattern (uniform or paroxysmal), the subjective qualitative character (aching, puncture, gravative, etc.), the qualitative-temporal character (non-incident or incident) and annotation of any complaints associated with the pain (Figure 4).

Pain felt deeply is undoubtedly tissue pain and cannot be neuropathic (with the exception of central pain), whereas pain felt superficially, if it is not superficial tissue pain from a skin or mucosal lesion, is definitely neuropathic.

Considering the criterion of intensity pattern, paroxysmal character (typical of trigeminal neuralgia) excludes all types of tissue and neuropathic pain with the exception of demyelination pain.

Within the criterion of subjective qualitative character, only the gravative-constrictive character that depicts deep tissue pain, the dysesthetic character that depicts a generic neuropathic pain, and above all the dysesthetic character that is strongly suggestive of demyelinating pain are relevant.

The criterion of the detectability of the algogenic lesion is relevant whether the lesion is at the site of pain denouncing superficial cutaneous-mucosal or deep musculoskeletal tissue pain and in relation to whether it is detectable at the site of pain (denouncing a primary pain) or elsewhere (denouncing a secondary pain).

The criterion of the presence of neurological deficits is relevant in that it eliminates the possibility of pain being tissue-based and points towards neuropathic pain, excluding pain from persistent nociceptor hyperexcitability, pain from demyelination and pain from small-fibre damage, and making axonal neuropathy pain or deafferentation pain (if it affects the trigeminal territory) highly probable.

Last, but of great importance, is the criterion of allodynia detection. The detection of superficial allodynia must lead one to assess the possibility of superficial cutaneous or mucosal tissue pain: if this possibility is excluded, it points towards pain from persistent nociceptor hyperexcitability, demyelination or axonal neuropathy, while excluding pain from small-fibre damage and deafferentation.

The finding of deep allodynia is obtained with the palpation part of the objective examination. Tender areas and myofascial trigger points should be considered deep allodynia. This finding should lead one to consider the possibility of visceral tissue, musculoskeletal or nerve trunk pain if it corresponds to a tender area and musculoskeletal tissue pain if it corresponds to a trigger point. The presence of a tender area must be considered in the case of pain caused by palpation of the abdomen in search of a visceral lesion, as well as pain caused by movement in the case of a musculoskeletal lesion (e.g. in coxarthrosis and gonarthrosis, supraspinatus tendonitis, etc.).

Once the pathogenetic diagnosis has been formulated, the algorithm will propose possible syndromic diagnoses (Figure 7) based on the site of the pain and the identification of the algogenic lesion, but it is the Algologist who selects the correct one.

With regard to syndromic diagnosis, it should be borne in mind that, although the primary objective of the algologist is to search for the pathogenesis of pain and in some circumstances this is sufficient to direct antalgic treatment, it may be important to identify the site of algogenic lesion in anticipation of a non-pharmacological but infiltrative or surgical treatment. For example, if one has recognised a deep musculoskeletal tissue pain felt in the upper limb in the deltoid region, instead of only proceeding with the appropriate treatment to counteract the pathogenic mechanism that sustains that pain, if one can identify the algogenic lesion in the supraspinatus tendon by completing the pathogenic diagnosis with the syndromic one, one can selectively act on that pathology. Considering other clinical situations, if in a patient with lumbar pain we reach the pathogenetic diagnosis of deep musculoskeletal tissue pain, we must complete the investigation to identify the site of the algogenic lesion because the treatment is different if it is at the level of an intervertebral disc, facet joint, ligament or muscular structures; in the case of an epigastric pain radiating to the upper lumbar region, which we identify as deep visceral tissue, the investigation must be completed with the search for the algogenic lesion that could be, for example, in pancreas and require not only antalgic treatment but also and above all direct treatment on the pancreatic lesion. Furthermore, it should be borne in mind that syndromic diagnosis is convenient for communication purposes.

Finally, the programme provides for a diary where subsequent algological visits are recorded for each patient with the therapeutic prescriptions and procedures performed, the assessment of pain intensity and the possibility of drawing up a clinical report.

Assuming that Algologist is not only the Pain Therapist and his task is not simply to perform pain therapy but first and foremost to make the pathogenetic and syndromic diagnosis and choose the treatment, the aspiring Algologist must have a set of knowledge derived, rather than from different specialisations, from medical knowledge in general that constitute what we can call the “cultural heritage of Algology”. Having reaffirmed the concept that “Algology” is the study of neuroanatomy, physiology and pathogenesis of pain and “Pain Medicine” (which the includes algological examination, the therapeutic decision and the execution of therapy) is the clinical application of Algology, the Algologist is the physician who knows Algology and is therefore able to practice Pain Medicine. It is on these premises that WinAlgos was conceived and realised, which started out as a simple Algological Clinical File and then evolved into a system that is intended to be a support for the formulation of the algological diagnosis.

At this point, one cannot help but draw attention to the much-fashionable Artificial Intelligence (AI), which has been very much in the news media in recent months, especially the “generative” one, i.e. the one capable of producing texts, images, videos and answers based on a question asked in the form of written text or voice recognition. What we are most interested in with reference to WinAlgos, however, is the part of artificial intelligence called ML (machine learning) and in particular the part called “deep learning” which deals with creating systems that learn or improve their performance based on the data used. Such techniques use artificial neural networks that attempt to replicate the behaviour of the human brain. This is actually an oversimplification since such systems are subject to technical constraints not present in biological systems. However, starting from large databases of specific data, supervised and validated by experts, the “artificial” neural network attempts to process real-time responses to user requests. In other words, this tool uses algorithms that, modelled on the decision-making processes of the human brain, can “learn” from the available data and make increasingly accurate predictions as the accuracy of the data increases. This is nothing new because for decades, computer science has been able to highlight connections and relationships in data at times that are difficult for humans to achieve. What has changed is the ability to process data in terms of hardware power (and thus processing speed) and above all the introduction of new algorithms based on the imitation of biological neural networks.

At this point, let us refer to what a doctor does every time he tries to make a diagnosis. Based on his experience, he assesses the symptoms, searches for possible inconsistencies in both the symptoms and the available instrumental inadequacies, and begins to work out the possible diagnosis. The more experience the doctor has, i.e. the more data he has memorised during his professional life (including any mistakes made in previous diagnoses), the more correctly he will be able to establish the diagnosis and treatment. Thinking now of being able to multiply the experience of the individual doctor by a large number of doctors willing to provide feedback on their experiences by storing them in a device such as WinAlgos, one realises how powerful and useful such technology could become.

For example, if one tries to use the free version of “Copilot”, one is initially surprised to get answers that are interesting at first glance, but then, going into detail, one finds that (even if they are very well summarised) they are absolutely generic and approximate, which the layman can get from internet searches... and that can only be so, since that is the data base for processing!

In fact, in order to get correct and really useful answers, the basis of the elaborations must be as reliable and validated as possible. At the moment, for the Algologist, the possibility of having such a tool to aid diagnosis seems a long way off, but in other areas of medicine (perhaps more codified and better funded) it is no longer just science fiction.

While waiting to organise a dedicated database, WinAlgos uses computer techniques that are more classical but not, if well managed, less interesting. While AI uses techniques that, simplifying, we could call statistical, WinAlgos for now uses procedural and deterministic techniques in the sense that the answers are uniquely determined by tables and algorithms used to generate the diagnosis. For example, for the formulation of the pathogenetic diagnosis, WinAlgos compares the evaluation criteria summarised in the table in Figure 6 with the findings in a series of tables, which should be further implemented as the programme is used by as many healthcare professionals as possible, thus becoming a system that is continually being improved.