Post

Designing a Dental Identification System: From Clinical Records to Structured Data

Designing a Dental Identification System: From Clinical Records to Structured Data

Designing a Dental Identification System

From Clinical Records to Structured Data

During my service as a public health dentist in Korea, I became increasingly aware of a structural limitation within forensic identification workflows: the way dental data is recorded and stored.

Dental structures are remarkably durable and often survive conditions that destroy other biometric identifiers. Because of this, dental records frequently play a critical role in identifying unknown individuals.

However, despite their importance, many dental records remain fragmented, unstructured, and difficult to process computationally.

This observation eventually led me to begin building a mobile-based dental identification system.


The Problem: Clinical Records Are Not Structured Data

Modern dental records contain large amounts of information.

Typical records include:

  • tooth-level conditions
  • restorations and materials
  • prosthetic work
  • implants
  • periodontal findings
  • radiographic observations

However, most of this information is recorded in formats optimized for human interpretation rather than computational analysis.

For forensic investigators performing identification, this creates a practical challenge. The process of comparing dental records often relies heavily on manual interpretation rather than systematic data comparison.

The issue is not the absence of information — it is the absence of structured representation.


Antemortem and Postmortem Comparison

Forensic dental identification primarily relies on comparing two datasets:

Antemortem (AM)
Dental records collected while a person was alive.

Postmortem (PM)
Dental findings collected from unidentified remains.

Identification occurs when patterns within these datasets align.

Typical comparison points include:

  • presence or absence of teeth
  • restorations and restorative materials
  • crown morphology
  • prosthetic structures
  • radiographic features

Even small consistencies across multiple teeth can significantly increase identification confidence.

However, efficient comparison requires the data to be structured in a consistent and machine-readable way.


Building a Structured Data System

To explore this problem from an engineering perspective, I began developing a prototype application designed to structure dental findings into machine-readable formats.

The system currently consists of:

Frontend

Flutter-based mobile interface designed for rapid data entry and visualization.

Data Layer

A structured dental schema encoding tooth-level information, surface findings, and prosthetic structures.

Storage

Firebase-based backend used to store and retrieve identification records.

Instead of storing dental information as free-text notes, the system records findings through structured data fields.

Examples include:

  • tooth status
  • restoration type
  • restoration surface
  • prosthetic indicators
  • missing teeth patterns

By structuring data at this level, the system becomes capable of supporting systematic comparison and future computational analysis.


Why Data Structure Matters

Forensic identification is fundamentally a pattern matching problem.

The reliability and efficiency of that process depends heavily on how well the underlying data is structured.

Once dental findings are encoded as structured datasets, it becomes possible to explore computational approaches such as:

  • similarity scoring between AM and PM datasets
  • automated pattern detection
  • integration with radiographic analysis systems

In many cases, the first step toward building intelligent systems is not building complex algorithms — it is structuring the data correctly.


Looking Forward

This project began as a practical attempt to improve how dental identification data is recorded and retrieved.

Over time, it has evolved into a broader exploration of how structured medical datasets can support intelligent systems.

As I continue studying software engineering and artificial intelligence, I am particularly interested in how data architecture and machine learning can support high-stakes domains such as forensic identification.

Reliable systems in these environments require more than algorithms.
They require carefully designed data structures that reflect real-world clinical information.

Building those foundations is where engineering begins.

This post is licensed under CC BY 4.0 by the author.