A review on cephalometric landmark detection techniques.

cephalo.png
Pastel Bake Sale Charity Events Website (20).png

Context

The cephalometric analysis investigates the relationship between the dental and skeletal features of the human skull. The analysis is frequently used as a treatment planning tool by dentists, orthodontists, oral and maxillofacial surgeons. The analysis relies on cephalometric radiography(i.e. X-rays of the human skull) to analyze relationships among bony and soft tissue landmarks. These landmarks are readily recognizable points on the radiographs and represent specific soft or hard tissue anatomical structures. The manual tracing of these cephalometric landmarks is a tedious and laborious task and prone to human errors, necessitating a need to develop efficient automated methods for landmark identification.

Application and Advantages

  • The cephalometric landmark tracing finds applications in diagnosing facial growth abnormalities and evaluating the treatment progress and results.

  • The study was performed to assist the practitioners and the researchers of this field to select the most compatible and appropriate technique for landmark tracing amongst the mentioned methods/software's to save on both time and energy consumed in comparison of various techniques in this field. The study included the following topics.

 
cephalo_topics.png
 

Methodology

This study reviewed and critically analyzed various techniques used for cephalometric landmark identification present in the current literature. We also tried to highlight the existing applications, address gaps in the literature, and present the open challenges of the field of study.

The study covered a wide range of techniques used for cephalometric landmark analysis, including knowledge-based, machine learning and deep learning-based techniques (depicted below).

cephalo_abstract.png

Result

The advent of artificial intelligence and machine learning has made the automation of cephalometric landmarking seemingly possible.

The study showed a general shift in trend from statistical and machine learning-based techniques to deep learning and convolution-based systems. Techniques such as active shape modelling, active appearance modelling, random forest regression-voting, Convolutional Neural Network (CNN), fuzzy systems and many others have rendered promising results in this field.