100% free

your diagnosis in your hands

marks the 25 cephalometric landmarks for you in seconds. review, adjust, save. export the report.

start free

everything you need in a cephalometry

fine adjustment

move each point with sub-pixel precision. keyboard arrows for fraction-by-fraction tweaks.

12 classic analyses

steiner, downs, tweed, eastman, wits, usp interlandi/vigorito/unicamp, mandibular, vertical, dental, basic.

build your own analysis

create custom analyses with your own points and metrics. save, export, import across accounts.

cervical maturation (cvm)

cervical stage (cs1-cs6) and pubertal growth spurt phase — computed automatically.

photo overlay

automatically overlays the patient's profile photo on the soft tissue of the radiograph. fully adjustable.

pdf report export

report with your brand, angles and measurements — ready to share with the patient.

automatic detection

25 cephalometric landmarks placed by ai in seconds. clinical-grade precision.

see it in action

referências

the science behind

studies and foundations that ground ezceph's methodology

[1]
AARIZ: A Benchmark Dataset for Automatic Cephalometric Landmark Detection and CVM Stage Classification
Khalid et al., 2025 · Scientific Data (Nature)
[2]
Deep High-Resolution Representation Learning for Human Pose Estimation (HRNet)
Sun et al., 2019 · CVPR
[3]
Cervical Vertebral Maturation as a Biologic Indicator of Skeletal Maturity
Baccetti, Franchi & McNamara, 2005 · Seminars in Orthodontics
[4]
Reliability of Cervical Vertebral Maturation Staging
Shimamura et al., 2024 · Scientific Reports
[5]
Cephalometric Analysis (Interlandi)
Interlandi S. · Ortodontia: Bases para a Iniciação, 5th ed., Artes Médicas, 2002
[6]
Cephalometrics for You and Me (Steiner)
Steiner CC. · American Journal of Orthodontics, 1953
[7]
cephalometric — open source web cephalometric analysis tool
Ali Saghiran (alexcorvi) · GitHub · MIT — base of our manual annotation SPA

frequently asked

the cephalometric landmark detector is an HRNet-W32 trained on a global base (AARIZ + a Brazilian dataset). it reaches a mean radial error (MRE) of around 2.4 mm on bone landmarks in the validation set — comparable to the inter-rater variability between experienced orthodontists annotating the same radiograph. soft tissue landmarks are tighter (~1.8 mm). always review before making clinical decisions — the ai assists, it doesn't replace.

curioso pra entender como tudo isso foi feito?

se rolar interesse o suficiente, a gente grava uma aula aberta mostrando a arquitetura, treino do modelo, decisões clínicas — tudo.

— pessoas querem

made by ezaligner

ezceph is provided by ezaligner — a brazilian aligner manufacturer passionate about technology. proud creators of the pit stop methodology and a unique way of treating patients.

learn about ezaligner