A framework for clustering dental patients' records using unsupervised learning techniques

This paper presents a method for hidden pattern mining on dental medical records related to oral conditions and different procedures that are performed on various patients. The decision to follow a set of procedures is based on the examination and diagnostics. Nowadays there is an increasing trend towards digital dentistry, but the full potential of digital data is not yet exploited because of several reasons. These reasons include technical issues like heterogeneous data gathering, storage strategies, restricted access or limited patient data and lack of expert systems over the gathered information. The paper addresses some of these issues and proposes a way to handle dental medical records. Data mining operations are performed to extract valuable rules and various interesting patterns along with their key indicators. No such work is yet performed in Pakistan in the field of dentistry. To this end, we analyze a real example of a dental hospital from Pakistan that treats patients (diagnosis and processes). Also, it is obvious that dental processes are far complex, which require an extensive domain expert knowledge for their manipulation. We use unsupervised learning techniques to perform clustering in order to discover interesting patterns that help us to determine the state of a patient at a particular instance by assigning them a label class. These patterns will also be helpful to medical practitioners to treat their patients wisely that what problems a particular patient may face. Research shows that most of the patients belong to mild and moderate dental patient's class. The most common problem that is being noticed in patients is tooth cavity with a treatment named “resin-based composite — one surface, posterior”. Also, we apply other well-known clustering algorithms on our dataset to analyze performance measures.