Transcript Biomedical images processing and analysis

Biomedical images processing and analysis
Biomedical images processing and analysis
Group members
Massimo De Luca (2003-)
Marco Foracchia (-2003)
Alfredo Giani (2003-)
Enrico Grisan
Lorenzo Marafatto (2005-)
Alfredo Ruggeri
Fellowship – PhD student
PhD student
Post doc
PhD student – Post doc
Fellowship
Associate professor
Biomedical images processing and analysis
Cooperations
J. Jaroszewski - Cornea Bank Berlin, Clinic of Ophthalmology, University School
of Medicine, Berlin, Germany
A. Neubauer - Dept. of Ophthalmology, Ludwig Maximilians University, Munich,
Germany
S. Piermarocchi – Dept. of Ophthalmology, University of Padova
D. Ponzin - Veneto Eye Bank Foundation, Venice, Italy
A. Pocobelli - Eye Bank, S. Giovanni-Addolorata Hospital, Rome, Italy
P. Gain - Ophthalmology Department, Bellevue Hospital, Saint-Etienne, France
A. Bezerianos - Dept. of Medical Physics, University of Patras, Greece
G. Barbaro - Nidek Technologies, Padova, Italy
P. Favaro - Siemens Corporate Research, Princeton (NJ), USA
Biomedical images processing and analysis
Publications
8
7
International Journals
Conference Papers
6
Conference Abstracts
5
4
3
2
1
0
2001
2002
2003
2004
2005
Biomedical images processing and analysis
Funding
University of Padova: € 60.000 (shared)
University of Padova: € 15.000
Ministry of University: € 20.000
Nidek Technologies: € 25.000
CARIPARO Bank Foundation: € 40.000
Biomedical images processing and analysis
1.Cell contour recognition for in-vivo microscopy
of corneal endothelium
Cell contour recognition
Statistical Correction
ROI extraction
Band-Pass Filtering
ANN contour extraction
Hole removal and Erosion
Skeletonization
Contour completion
Perimeter extraction
Correction
Cell contour recognition
Statistical Correction
ROI extraction
Band-Pass Filtering
ANN contour extraction
Hole removal and Erosion
Skeletonization
Contour completion
Perimeter extraction
Correction
Cell contour recognition
Statistical Correction
ROI extraction
Band-Pass Filtering
ANN contour extraction
Hole removal and Erosion
Skeletonization
Contour completion
Perimeter extraction
Correction
Cell contour recognition
Statistical Correction
ROI extraction
Band-Pass Filtering
ANN contour extraction
Hole removal and Erosion
Skeletonization
Contour completion
Perimeter extraction
Correction
Cell contour recognition
Statistical Correction
ROI extraction
Band-Pass Filtering
ANN contour extraction
Hole removal and Erosion
Skeletonization
Contour completion
Perimeter extraction
Correction
Cell contour recognition
Statistical Correction
ROI extraction
Band-Pass Filtering
ANN contour extraction
Hole removal and Erosion
Skeletonization
Contour completion
Perimeter extraction
Correction
Cell contour recognition
Statistical Correction
ROI extraction
Band-Pass Filtering
ANN contour extraction
Hole removal and Erosion
Skeletonization
Contour completion
Perimeter extraction
Correction
Cell contour recognition
Statistical Correction
ROI extraction
Band-Pass Filtering
ANN contour extraction
Hole removal and Erosion
Skeletonization
Contour completion
Perimeter extraction
Correction
Cell contour recognition
Statistical Correction
ROI extraction
Band-Pass Filtering
ANN contour extraction
Hole removal and Erosion
Skeletonization
Contour completion
Perimeter extraction
Correction
Nidek Technologies NAVIS-ENDO system
•
The ENDO software is a module of the
system for ophthalmology.
A glimpse of tomorrow …
Biomedical images processing and analysis
2.Fourier analysis for the estimation of cell density
on eye bank images of donor corneas
Fully automatic density estimation
AIM:
• to develop a fully automatic technique for cell density
estimation (no user intervention).
• It must be without cell contour detection.
2250 cell/mm2
Frequency-based density estimation

A repetitive pattern of cell
borders is clearly visible.

Spatial frequency of this
pattern is proportional to
cell density.

Frequency information is
available through Fourier
analysis.

Information from Fourier
analysis can provide an
estimation of cell density.
Frequency-based density estimation
Gray-scale image of 2D-DFT log-magnitude.
A circular band
indicates that the
endothelium
image contains a
repetitive pattern
at a specific
frequency.
Spatial frequency is the radius of the band
Radius of circular band
can be used to estimate
cell density.
Frequency-based density estimation
Position of second peak provides estimated frequency f of cell
borders.
Frequency-based density estimation
Automatic density (cell/mm2)
3500
3000
2500
2000
1500
1500
2000
2500
3000
3500
2
Manual density (cell/mm )
(Ruggeri et al., Br J Ophthalmol, Mar 05)
Nidek Technologies NAVIS-EyeBank system
•
The EyeBank software is a module of the
system for ophthalmology.
Biomedical images processing and analysis
3. Tracking techniques for vessel-like structure
Applications to:
• vessels in retina
• nerves in cornea
Clinical outcomes:
• length
• tortuosity
• bifurcations
• caliber course
• optic disc detection
Tracking techniques in retina
Tracking techniques in retina
Tracking techniques in cornea
Biomedical images processing and analysis
4.Methodologies in eye fundus analysis for the
diagnosis of retinopathy
Diabetic retinopathy characterized by fundus lesions
Automatic and objective tools:
• wide screening
• disease assessment &
monitoring in time
• (new) drugs efficacy
Eye fundus analysis
Three steps:
1. Detection
2. Classification &
measurement
3. Clinical
assessment
Biomedical images processing and analysis
5.Design and realization of an adaptive optics
fundus camera
Eye
Retinal
Imaging
Flash
path
Wavefront
sensor
Image Processing
Biomedical images processing and analysis
5.Design and realization of an adaptive optics
fundus camera
Image
acquisition
Coma
Mirror
update
Image
Analysis
Defocus
Astigmatism
Image
corrected
Biomedical images processing and analysis
6.Bio-inspired omni-directional vision
Omnidirectional mirror
Development of
biomimetic
algorithms
for vision
Space variant sensor
providing with a
foveated vision