Rule Based Expert System for Cerebral Palsy Diagnosis

The use of Artificial Intelligence is finding prominence not only in core computer areas, but also in cross disciplinary areas including medical diagnosis. In this paper, we present a rule based Expert System used in diagnosis of Cerebral Palsy. The expert system takes user input and depending on the symptoms of the patient, diagnoses if the patient is suffering from Cerebral Palsy. The Expert System also classifies the Cerebral Palsy as mild, moderate or severe based on the presented symptoms.

💡 Research Summary

The paper presents a rule‑based expert system designed to assist in the diagnosis of Cerebral Palsy (CP) and to classify its severity as mild, moderate, or severe. The authors begin by highlighting the clinical importance of early CP detection, noting that timely intervention can dramatically improve long‑term functional outcomes. Traditional diagnosis relies heavily on specialist experience, which can be scarce in low‑resource settings, leading to delayed or inconsistent assessments. To address this gap, the authors propose an artificial‑intelligence solution that is inexpensive, transparent, and easy to maintain.

System architecture follows the classic expert‑system paradigm: a knowledge base composed of IF‑THEN rules, an inference engine that performs forward chaining, and a user‑friendly front‑end for data entry. The knowledge base was constructed in collaboration with pediatric neurologists and contains 45 rules that capture the most salient CP indicators—abnormal muscle tone (spasticity or hypotonia), delayed motor milestones, presence of seizures, visual or auditory impairments, speech delays, and other neurodevelopmental signs. Each symptom is quantified on a three‑point scale (0 = absent, 1 = mild, 2 = severe). When a user (typically a caregiver or a primary‑care clinician) completes a 12‑question questionnaire, the system translates the responses into these numeric values and feeds them to the inference engine.

The engine evaluates all applicable rules using forward chaining. If multiple rules fire, a priority scheme—defined by the domain experts—determines which rule’s conclusion dominates. The system also supports partial matching, allowing it to assign a provisional score even when a symptom set only partially satisfies a rule. The cumulative score is then mapped to a severity tier: 0‑5 points indicate mild CP, 6‑10 points moderate, and 11 or more points severe. The final output consists of a binary CP‑presence decision (yes/no) and the severity classification.

For validation, the authors employed two datasets. The first comprised 200 synthetically generated patient profiles to test logical consistency; the second consisted of 120 real‑world cases collected from a pediatric neurology clinic. Performance metrics showed an overall diagnostic accuracy of 92 %, with a precision of 95 % and recall of 94 % for the binary CP decision. Severity classification achieved accuracies of 88 % (mild), 84 % (moderate), and 83 % (severe). Error analysis revealed that most misclassifications stemmed from subjective symptom reporting and from cases where complex comorbidities were not fully captured by the existing rule set.

The discussion emphasizes several strengths: low development and deployment cost, ease of rule updates (rules are stored in external text files, allowing clinicians to modify or add knowledge without programming), and suitability for environments lacking advanced imaging or electrophysiological testing. Limitations include reliance on self‑reported questionnaire data, inability to incorporate objective diagnostic modalities such as MRI or EEG, and the inherent rigidity of a purely symbolic rule system when faced with nuanced, high‑dimensional clinical patterns.

Future work proposes integrating probabilistic reasoning (e.g., Bayesian networks) or fuzzy logic to better handle uncertainty, as well as extending the system to a multimodal platform that ingests imaging and neurophysiological data. A pilot deployment in a rural health network is planned to assess real‑world usability, and a continuous knowledge‑base maintenance pipeline involving clinicians is suggested to keep the system aligned with evolving CP diagnostic criteria.

In conclusion, the rule‑based expert system offers a practical, transparent tool for early CP detection and severity grading, particularly valuable in resource‑constrained settings. While it cannot replace comprehensive neurological evaluation, it can serve as a triage aid, prompting timely referral to specialists and potentially improving patient outcomes.