Software Reuse in Medical Database for Cardiac Patients using Pearson Family Equations

Software reuse is a subfield of software engineering that is used to adopt the existing software for similar purposes. Reuse Metrics determine the extent to which an existing software component is reused in new software with an objective to minimize the errors and cost of the new project. In this paper, medical database related to cardiology is considered. The Pearson Type I Distribution is used to calculate the probability density function (pdf) and thereby utilizing it for clustering the data. Further, coupling methodology is used to bring out the similarity of the new patient data by comparing it with the existing data. By this, the concerned treatment to be followed for the new patient is deduced by comparing with that of the previous patients case history. The metrics proposed by Chidamber and Kemerer are utilized for this purpose. This model will be useful for the medical field through software, particularly in remote areas.

💡 Research Summary

The paper proposes an integrated framework for reusing software components within a cardiac‑patient medical database, aiming to reduce development cost, minimize errors, and accelerate treatment decisions, especially in remote healthcare settings. The authors begin by preprocessing a large collection of cardiology records, normalizing clinical variables such as blood pressure, heart rate, cholesterol levels, and ECG parameters. They then model the probability density function (pdf) of each variable using the Pearson Type I distribution, a member of the Pearson family capable of representing asymmetric and heavy‑tailed data through its four moments (mean, variance, skewness, kurtosis). This choice is justified because cardiac data often deviate from normality, and Pearson Type I provides a flexible fit without resorting to mixture models.

With the pdfs in hand, the authors apply an Expectation‑Maximization (EM) clustering algorithm to uncover latent patient groups. Each cluster is characterized by a centroid and a covariance matrix, and the assignment of a new patient to a cluster is performed by evaluating the posterior probability under the fitted Pearson distributions. To capture the asymmetry of the underlying data, they employ a modified Kullback‑Leibler divergence as the distance metric, ensuring that the most statistically similar cluster is selected.

The second major contribution lies in the software engineering layer. The system is decomposed into modular, object‑oriented components that handle data ingestion, pdf calculation, clustering, similarity matching, and treatment recommendation. To quantify the quality of these components, the authors adopt the Chidamber‑Kemerer (CK) suite of metrics—CBO (Coupling Between Objects), DIT (Depth of Inheritance Tree), NOC (Number of Children), RFC (Response For a Class), and LCOM (Lack of Cohesion of Methods). High CBO and RFC values indicate excessive inter‑module dependencies, while a high LCOM signals low internal cohesion; both conditions are detrimental to reuse. The paper provides concrete refactoring guidelines: for instance, if CBO exceeds ten, extract interfaces; if LCOM surpasses 0.8, split the class into more cohesive units. By continuously monitoring these metrics, the architecture can be kept lean, maintainable, and amenable to reuse.

The third pillar of the framework is a coupling‑based similarity analysis that bridges statistical clustering with software metrics. For each new patient, a similarity score is computed against records in the assigned cluster. The score aggregates variable‑level weights (e.g., blood pressure and heart rate receive higher clinical importance) and CK‑derived weights (e.g., lower CBO contributes positively). When the composite similarity exceeds a predefined threshold (e.g., 0.75), the system retrieves the treatment protocol associated with the most similar historical case and presents it as a recommendation. This approach effectively translates statistical similarity into actionable clinical guidance while leveraging the software quality indicators to ensure that the underlying code remains robust.

Finally, the authors address deployment in low‑bandwidth, remote environments. Rather than transmitting full patient histories, the server sends only cluster centroids and a compact summary of CK metrics. The client, running a lightweight inference engine, matches the incoming patient data to the nearest centroid, computes the similarity score, and instantly displays the suggested treatment pathway. This design dramatically reduces network traffic, enables near‑real‑time decision support, and makes the system viable in underserved regions where traditional telemedicine infrastructure is limited.

In summary, the paper presents a novel synthesis of statistical modeling (Pearson Type I distribution), object‑oriented reuse metrics (Chidamber‑Kemerer), and coupling‑based similarity analysis to create a reusable, scalable, and clinically useful software platform for cardiac patient management. By grounding treatment recommendations in both data‑driven clusters and rigorously measured software quality, the proposed model promises to improve diagnostic accuracy, lower development overhead, and extend advanced cardiac care to remote populations.