Post

Querying Dental Data for Identification: From Static Records to Computable Intelligence

Querying Dental Data for Identification: From Static Records to Computable Intelligence

Introduction

In forensic dentistry, data collection is only half of the problem.
The real challenge begins when we try to use that data for identification.

Traditional dental records—charts, notes, radiographs—are designed for human interpretation.
They are inherently static.

However, in real-world scenarios such as mass disasters or unidentified remains, this approach quickly becomes inefficient.

The critical question is:

How can we transform dental records into a system that is queryable, comparable, and computationally meaningful?

This post focuses on that transformation:
querying dental data for identification.


Limitations of Traditional Workflows

In most current workflows:

  • Dental findings are manually recorded
  • Comparisons are performed visually
  • Matching depends heavily on expert experience

This leads to fundamental limitations:

1. Poor Scalability

Comparing dozens or hundreds of cases becomes impractical.

2. Inconsistency

The same clinical finding may be described differently across practitioners.

3. Lack of Queryability

There is no structured way to ask:

  • “Find all cases with a missing upper left first molar”
  • “Match this restoration pattern”

In short:

The data exists, but it is not computable.


Designing a Queryable Data Structure

To address this, I structured the system around a single principle:

Every dental observation must be stored in a machine-readable, queryable format.


1. Tooth-Level Indexing (FDI System)

Each tooth is indexed using the FDI numbering system (1–32),
providing a consistent coordinate system across all records.


2. Surface-Level Representation

Each tooth contains surface-specific data:

  • O (Occlusal)
  • M (Mesial)
  • D (Distal)
  • B (Buccal)
  • L (Lingual)

This enables fine-grained queries such as:

  • Comparing occlusal restorations
  • Identifying mesial surface treatments

3. Global Attributes

Each tooth also includes global attributes:

  • Crown status
  • Root condition
  • Position
  • Pathology
  • Bite/Occlusion

These fields are essential for higher-level comparisons.


4. Hierarchical Code System

Free-text input introduces inconsistency.
To prevent this, all inputs are constrained through a hierarchical code system.

  • Users select from a structured tree
  • Each selection is stored as a path

This ensures:

  • Standardization
  • Machine interpretability
  • Future compatibility with AI models

From Data to Query

Once structured, the data becomes actionable.


Exact Matching

Example:

  • Tooth 11 has a crown
  • Tooth 26 is missing

This can be resolved through direct filtering.


Pattern Matching

More advanced cases involve patterns:

  • Distribution of restorations across quadrants
  • Symmetry between left and right arches

This transforms the problem into:

  • Feature extraction
  • Similarity comparison

Partial Matching

Real-world data is often incomplete.

  • Missing teeth
  • Partial records

This requires:

  • Flexible query logic
  • Weighted scoring
  • Tolerance for missing data

Why This Matters

This transition fundamentally changes the system:

  • Identification becomes faster
  • Large datasets become manageable
  • Matching can be assisted or automated

Most importantly:

The system evolves from documentation to computation.


Toward AI-Assisted Identification

Once the data is structured and queryable:

  • Pattern-based models can be trained
  • Similar cases can be suggested
  • Identification can be partially automated

Without structured data, none of this is possible.


Conclusion

The key insight is simple:

Dental data only becomes valuable when it is queryable and comparable at scale.

Building such a system is not just about recording information.
It is about transforming it into something that can be computed, searched, and reasoned about.

This is where software engineering meets forensic science.

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