Content uploaded by Eric Wisotzky
All content in this area was uploaded by Eric Wisotzky on Jul 30, 2019
Content may be subject to copyright.
Abstract— We present a stereo-endoscopic framework that
allows real-time capable image-based measurements of
anatomical structures. A 3D endoscope is used to reconstruct the
oral cavity to measure orofacial cleft of five pediatric patients.
Orofacial cleft is a congenital birth defect. It often comes
with several physical and aesthetic problems for the patient
including breathing difficulties, speech problems, food intake,
dental malposition or negative influence of middle ear
ventilation. To overcome these problems, it is necessary to
perform a surgery. For a cleft palate intervention, an obturator
plate is needed to cover the cleft. This plate is created by a
silicone impression while the patient is under anesthesia. It is
desirable to avoid such interventions by replacing the
impression by an optical measurement to reduce surgical risk
and improve patient’s well-being.
For a correct and robust 3D reconstruction, a calibration of
the stereoscopic system is crucial to get a real world
representation of the surgical scene and the underlying optical
system. The calibration method follows a model-based
approach using synthesized images and applying an image
registration via gradient-descent. For an overview of
calibrating medical imaging systems, we refer to [1,2]. A
successful calibrated stereo system allows performing
image-based measurements. Therefore, a real-time capable
image processing pipeline is applied, which has been used for
simple point-to-point distance measurements inside the
tympanic cavity . This measurement tool consists of two
steps: It performs a quasi-auto calibration by a scene
dependent rectification of image pairs and a highly
parallelized sub-pixel dense disparity estimation, reducing the
correspondence problem to a 1D-search. This results in a fast
and robust stereo estimation of correspondences to calculate
the related metric depth information. This depth information
allows an accurate measurement of the captured anatomical
structure . For this work, five patients with lip/cleft palate
are evaluated. The cleft size is measured using the
point-to-point measurement tool. In addition, the oral cavity
including the relevant cleft palate is 3D reconstructed.
For all five patients, the oral cavity including lip cleft and
palate cleft could be reconstructed. Fig. 1 shows
reconstruction results of the oral cavity for two patients and
the survey of the cleft palate in two directions. Besides the
*Research supported by the BMBF under grant number 16SV8061.
J.-C. Rosenthal, E.L. Wisotzky, P. Eisert are with Fraunhofer HHI, e-mail:
email@example.com. F.C. Uecker is with Charité Berlin
successful reconstruction, we can also perform valuable
image-based measurements of the cleft palate to determine its
size. For further validation of our results, the 3D endoscope
has been validated using specimens with known dimensions.
Empirical specimen evaluation give accuracies of approx.
1/10mm for several hundred point-to-point measurements.
Figure 1. Two single-shot 3D reconstructions results of oral cavity. Upper
row: point-to-point distance measurements of cleft: 4.3mm/22.3mm. Bottom
left: several ground-truth measurements on specimen.
IV. DISCUSSION & CONCLUSION
Based on this true-scale anatomy representation, the surgeon
will be able to measure the exact size and extract the anatomy
without the need of a silicone impression and several fitting
iterations. From this approach, three main benefits arise: (1)
the risk of complications caused by harmful tissue contact is
reduced, (2) an exact patient anatomy is received in near
real-time and (3) anesthesia can be avoided. As next steps, we
focus on the following two issues: (1) comparing our
reconstruction results to existing obturator plates (2) planning
an easy-to-use measurement toolbox for the surgeon.
We thank Schölly Fiberoptic GmbH, Denzlingen, Germany
for providing a 3D endoscope. Charité – Universitätsmedizin
Berlin ethically approves all measurements.
 N. Gard et al. Image-based Measurement by Instrument Tip Tracking
for Tympanoplasty using Digital Surgical Microscopy, Proc. SPIE
10951, Medical Imaging 2019.
 J.-C. Rosenthal et al. Kalibrierung stereoskopischer Systeme für
medizinische Messaufgaben, Proc CURAC, pp. 159-161, 2017.
 J.-C. Rosenthal et al. Microscopic image-based determination of
stapes prosthesis length. Proc. CARS, pp. 59–60, 2018.
 E. L. Wisotzky et al. Interactive and Multimodal-based Augmented
Reality for Remote Assistance using a Digital Surgical Microscope.
AVEH Workshop at IEEE VR 2019.
Endoscopic Single-shot 3D Reconstruction of Oral Cavity
Jean-Claude Rosenthal, Eric L. Wisotzky, Peter Eisert, and Florian C. Uecker