Prediction of COVID-19 based on neurological manifestations using a fuzzy logic system

The COVID-19 disease, caused by the Severe Acute Respiratory Syndrome coronavirus 2 (SARS-CoV-2) virus, becomes one of the most fatal diseases that have ever affected human life and global economies, which has been reported in ranges from mild to severe in clinical representation [1]. The disease does not only have respiratory symptoms but also shows symptoms related to many organ systems, including the nervous system [2]. Clinical reports showed that neurological symptoms had presented a significant frequency of extrapulmonary manifestations [3]. For example, headache is given between signs in 13.6% of total COVID-19 patients with range means ranging between 4 and 24%) [4]. In addition, myalgia has shown a frequency of 20.4%, with an average ranging between 11 and 44%. [5] Furthermore, anosmia has been reported in the range of 5–88% [6]. In addition, some studies reported the frequency of neurological manifestations in severe illness cases of COVID-19 patients as 84%, while it was 36.4% in hospitalized cases [7,8]. In these studies, neurological manifestations showed up more frequently in severe cases. However, the prognoses’ influences are still not apparent [9].

In addition, surveying dozens of research works in the literature [10‐20] showed a significant presence of neurological symptoms in COVID-19 patients, including mild and severe cases. For example, headache has been reported in 70–80% of all cases; anosmia–dysgeusia showed up in 50–60% of total cases; myalgia presented in 40–45% of all cases; fatigue was there in 30–35% of the surveyed cases; dizziness was recorded in 30–35% of patients; and 0–10% of subjects showed noncommon symptoms like numbness, migraine, loss of concentration, and seizure. The obvious correlation between some neurological symptoms and COVID-19 with variable ratio and weight can be utilized to predict the likelihood of having COVID-19 or not based on neurological symptoms. Therefore, this study analyzes thousands of COVID patients with neurological symptoms reported in the literature by surveying a couple of related research works to achieve twofold objectives. First, it determines statistically the weight of each neurological symptom in COVID-19 cases. Second, the obtained statistics of neurological manifestations are used to design the ruleset of a fuzzy logic-based model for predicting the likelihood of having COVID-19 based on neurological symptoms. To sum up, the survey of dozens of related research works yields thousands of records showing the neurological symptoms that appeared in COVID-19 cases. Then, the weight and percentage ratio of these symptoms are statistically calculated. Afterward, these data are utilized to form the ruleset of a fuzzy-logic-based system. As a result, the fuzzy logic system can decide the likelihood and certainty of having COVID-19 or not for patients based on their neurological symptoms.

Because the research planned to perform statistical analysis as a first stage and use its outcomes to build the rulesets of the fuzzy logic system to predict the likelihood of COVID-19 based on the neurological manifestations, the results are expected to be for the statistical and fuzzy logic results. Starting with the statistical results, the outcome, as represented in Fig. 3. depicts the neurological manifestations of COVID-19, showing headache is the highest percentage with 71.96. Anosmia and loss of taste come next with 52.34%, and all other symptoms are shown in the graph.

Bild vergrößern

Based on the results of the statistical analysis, as shown in Fig. 3, by applying the resulting values of the neurological manifestations of COVID in the predesigned fuzzy logic rule set shown in Fig. 4, the likelihood of COVID-19 based on the neurological symptoms can be determined. The shaded areas in yellow in Fig. 4 mean that the parameter is realized at that area. For example, to have the output that represents the likelihood of COVID-19 realized, the headache manifestation should be more meaningful than 50%, and loss of taste and anosmia be mild from 25 to 75%, as shown in the second row in Fig. 4.

Bild vergrößern

Two test cases are tested by the developed fuzzy logic system that forecasts the COVID-19 likelihood based on neurological manifestations. In the first scenario, one patient comes with a headache at a level of 50% and no other symptoms and the second case is for another patient who comes with a headache of 50% and loss of taste at a level of 80%. The system decided the likelihood of COVID-19 for the first patient to be zero or no COVID-19; for the second case, it decided the case as COVID-19 diagnosed. In addition, 20 test cases of the given data have been tested by the developed fuzzy logic. The results are reported in Table 3, showing the data of all symptoms, the results predicted by the developed fuzzy system, and the actual results based on clinical and laboratory investigations for validation. When the redevelopment test reports “yes,” that means it is confirmed covid by clinical and laboratory analysis or the likelihood of COVID-19 is true. But when the result states “NO,” that means no covid confirmed by laboratory or clinical investigation. The possibility of COVID-19 is reported false by the fuzzy system based on the given symptoms.

There are no clinical tests specific to COVID-19 apart from laboratory confirmation. Under pandemic conditions, complete blood count, CT imaging, pneumonia specific to COVID-19, and saturation values are predictive values in cases of severe disease. Nevertheless, the known symptoms are not typical for achieving a reliable diagnosis of non-severe cases of COVID-19 [21]. At the same time, the testing capability under real conditions is not enough to meet current needs. Therefore, forming target groups for self-isolation and outpatient treatment is an important issue . [22]

Since the first series of COVID-19 cases, neurological symptoms have been considered the most frequent extrapulmonary reported manifestations [2,3,5]. Neurological symptoms in COVID-19 patients generally arise between 1 and 14 days after the onset of SARS-CoV2 infectious symptoms [23].^.

In this study, we aimed to evaluate the diagnostic accuracy of neurological symptoms early in the course of the disease and their predictive values in patients with suspected COVID-19, which is expected to make a clinical diagnosis. Therefore, the data of thousands of COVID-19 patients, including their neurological symptoms, have been collected by surveying dozens of related works in literature. Then, we investigated the association between neurological manifestations in one of those studies10 in which 107 COVID-19 patients were self-isolated during the initial 10 days of disease onset to determine the type and frequency of symptoms in COVID-19 cases. After that, we applied the results of statistically analyzing the collected data to a developed fuzzy logic-based model aiming to forecast the certainty of the likelihood of COVID-19 in patients with neurological manifestations.

One hundred percent of patients present at least one neurological manifestation early in the disease. The most prevalent was headache which was reported by 72% of patients. The next was the loss of smell and taste which was reported as 52%. Then, muscle pain was reported at 44%, followed by fatigue at 33%, and dizziness at 32%. In addition to being frequent, the neurological symptoms had an early presentation. In addition, headache criteria were dull, aching, moderate intensity, like tension headache in most patients compared to migraine (67% vs. 10% migraine-like). In addition, two patients had occipital headache. Overall, none of the patients had a history of primary headaches. Hence, headaches in those patients could be attributed to SARS-CoV2 infection, not an aggravation of the previous primary headache. Anosmia and dysgeusia have arisen as the chief predictors of COVID-19. In a multicenter case–control study on 777 COVID-19 cases, anosmia and dysgeusia had higher odds ratios of 6.21 and 2.42, respectively, for positive SARS-CoV-2 infection. In addition, the application of machine learning algorithms achieved an accuracy of an average of 80% (82% sensitivity and 78% specificity), with anosmia and dysgeusia being the most prevalent symptoms to accurately predict COVID-19 [24].Even though the mechanisms are still not clear, correlation of anosmia linked with nasal obstruction or rhinorrhea, gives a clue for the importance of olfactory epithelium in COVID-19 [25‐27]. The overexpression of ACE2 receptors in the airway (the route facilitates SARS-CoV-2 entry into cells) in the olfactory neuroepithelium might be responsible for the nervous system manifestations reported in COVID-19 disease and the entry zone through which the virus reaches the central nervous system [28]. Furthermore, there might be other routes the virus uses to reach the brain, such as the gastrointestinal tract and respiratory system. 29 Besides, the eyes could be a site for the virus to reach the CNS through the trigeminal nerve [30] as evidenced by the presence of SARS-CoV-2 in tears and conjunctival secretions. [31]

Prediction models that combine several variables or features to estimate the risk of being infected with COVID-19 have been proposed as valuable tools to assist medical staff in triaging patients, especially when allocating limited healthcare resources [25]. For the validation of the fuzzy logic system, 20 cases of the original given data that have been used in this study were tested; these tests came up with matched decisions between the one taken based on clinical and laboratory investigation, which is represented by the 9th column in Table 4 and the results produced by the developed fuzzy logic which meant in the 12th a last column in the same table (Table 4). Statistical results confirmed the association between COVID-19 in its first week and neurological manifestations having different frequency rates for each symptom. The headache was the most significant frequency, and then anosmia and loss of taste came next. After that, myalgia and fatigue took the following order. The results showed that 19 out of 20 tests are matched between the lab results and the prediction results. That proved an accuracy of 95% of the designed system based on neurological manifestations and a fuzzy logic system in predicting COVID-19 diagnosis. Hence, increasing the awareness that COVID-19 patients may present with non-specific neurological manifestations among general practitioners, emergency department physicians, and neurologists is very crucial [32]. According to the results of the conducted survey of dozens of related research works, neurological symptoms have not yet been investigated regarding the prediction and diagnosis of COVID-19. So far, studies for predicting mild COVID-19 depend on non-specific manifestations: fever, cough, malaise, nasal obstruction, sore throat, headache, and myalgia [21]. These symptoms are typical for mild acute upper respiratory tract viral infections, particularly acute nasopharyngitis, caused by already known human coronaviruses [33]. Anosmia was proven to be a clinical marker for COVID-19 under pandemic conditions [34‐36]. In Germany, 2 of 3 confirmed cases are reported to have anosmia [37]. Popovich V and colleagues analyzed clinical symptoms' interaction and predictive importance to determine the diagnosis of mild COVID-19 using the ordinal regression method. Of the 120 patients who underwent testing, COVID-19 diagnosis was established in 96 patients and ruled out in 24. When assessing symptoms by a physician, according to the correlation analysis, fever, muscle pain, nasal congestion, and rhinolalia have a positive predictive value of more than 0.6 significance values. Nasal discharge, cough, and sore throat have negative predictive values [38]. However, most prediction models for COVID-19, according to a recent systematic review, were considered poorly reported, at high risk of bias, and probably report an over-optimistic performance [39]. Therefore, there are many limitations when transferring the prediction models into clinical practice, which cast uncertainty on their actual applicability, especially for different outbreak scenarios and populations which was similar in different applications that have used different prediction methods [40‐43].

Surveying data of neurological and non-neurological symptoms of thousands of COVID-19 patients in many related literature showed neurological manifestations in all patients with different ratios and weights, including mild and moderate cases. By statistically analyzing these data aiming to form the rulesets of a predesigned fuzzy logic-based forecasting system, the fuzzy logic system was able to yield a prediction of the likelihood of having COVID-19 in a group of patients based on their neurological symptoms with an accuracy of 95% by comparing the predicted data with the clinical data.