Transcript Observer Study of NeoTect SPECT

Observer Study of Reconstruction
Strategies for Detection of
Solitary Pulmonary Nodules
Using Hybrid NeoTect SPECT
Images
Xiaoming Zheng, PhD.
20 October, 2004
Outlines
•
•
•
•
•
•
•
Lung Cancer and SPECT/PET
NeoTect in Lung SPECT
Image Reconstructions: RBI vs FBP
Hybrid Images: Clinical Reality
Observer Studies: Human vs Numerical
ROC: Receiver Operating Characteristics
Results and Conclusions
Leading Causes of Cancer Deaths
180,570
Lung
Colon/Rectum
Breast
328,365
Stomach
Prostate
110,669
55,704
76,030
59,088
Others
NSCLC: Non-Small-Cell-LungCancer
• Surgery is providing the best chance of
cure if tumor can be re-sected completely.
• If cancer has spread to contra-lateral
lymph nodes or beyond the chest surgery
alone is not useful. Chemotherapy and/or
radiotherapy are usually applied. These
measures are rarely curative
SPN: Solitary Pulmonary Nodule
• Approx. 30% of new cases of lung cancer
are found as an SPN
• An SPN is defined as:
– single pulmonary lesion
– well defined borders
– mean diameter not more than 3 cm
• Found in 1 : 500 chest X-rays
SPECT and PET
(With chest XRay)
Patients
NeoTectSPECT
114
FDG - PET
Sensitivity
97%
98%
Specificity
73%
69%
Accuracy
91%
89%
89
NeoSpect
NeoTect/SPECT vs FDG/PET
P
L
P
P
FDG-PET
P
P
L
P
NeoTect (99mTc-Depreotide)
• Binds to Somatostatin receptors, which are overexpressed in lung cancer (NSCLC and SCLC)
• Has a negative predictive value of up to 98% in
combination with CT or chest X-ray for SPN
• Procedure is non-invasive
•
99mTc-labelled
- readily available
• Procedure is easy
• Can be used wherever SPECT is available
NeoTect
Binding
region for
SSTR*
NH2
NH2
(N-Me)Phe-Tyr
O
D-Trp
O
H2N
NH
O
N O N
Tc
S
H2N
O
Lys
Hcy-Val
NH
O
- a small synthetic peptide
- 10 amino acids, mol. wt. 1358 Da
- binding region for the somatostatin receptor
- radio-labeled with 99mTc
How NeoTect Works
–Malignant tumors over-express
somatostatin receptors (SSTRs)
–NeoTect binds to and detects
SSTRs
–Most benign lesions do not overexpress SSTRs
Normal Transaxial SPECT Images
Coronal SPECT
72 yr female smoker, complaining of
weight loss; chest x-ray: 2.5 cm LUL
lesion; CT: LUL 2.0 cm spiculated
mass; Histopathology (CT guided
FNA biopsy): poorly-differentiated
adenocarcinoma
Transaxial SPECT
CT
Aims of This Work
• Use hybrid images of lung tumor imaging
agent Tc-99m NeoTect in Localization
Receiver Operating Characteristic (LROC)
studies to determine reconstruction
parameters and whether iterative
reconstruction with attenuation, scatter,
and distant resolution compensation
should replace FBP clinically.
Why Hybrid Images
• The Optimization of reconstruction parameters,
and determination of whether iterative
reconstruction should replace FBP clinically
should be based on tasks which closely
approximate the clinical application of the
images
• The use of hybrid images or studies represents
a practical alternative to the use of purely clinical
acquisitions for observer studies.
How Hybrid Images Were Created
• Simulated lesions are added to know normal
clinical acquisitions
• Monte Carlo simulation package SIMIND was
used to simulate lesions.
• Nine normal patient’s projection data were used
to create 162 tumors randomly distributed within
the lung regions.
• Tumors were 1 cm in diameter which is the
smallest tumor could be detected by CT.
NeoTect Projections
From Clinical 9 Patients
Tumor Source Projections
From Monte Carlo Simulation
Images Reconstructions
• Iterative Reconstruction: Rescaled Block
Iterative Algorithm including attenuation, scatter,
and distance resolution compensation.
Parameters tested: iteration 1,3,5,7,10 and post
Gaussian filter FWHM 0,1,2,3,4 pixels
• Filtered Back-Projection: Parameters tested:
Butterworth filter cut-off frequencies: 0.10, 0.15,
0.20, 0.25 and 0.30 pixel-1
Filtered Back-Projection
Butterworth Filter and Cutoff
Frequency
FBP Reconstructed Images
Iterative Reconstruction
Rescaled Block Iterative
Reconstruction Algorithm
fn
k 1
k

c rn 
fn
 f 1 

a r  a r  H mn

k
n
m
c rn
d m  S 'm
H mn

Hf m
m S

k
r
  H mn 
 m S r

 
; a r  max c rn 

H mn 

 m

Attenuation Compensation
Scatter Compensation
 C1   C 3  

S '  .5W 2     
  W1   W 3  
Resolution Compensation
RBI Reconstructed Images
Receiver Operating Characteristics
Images for Observers
Numerical Observers
Types of Channels
Human
Observer
Interface
RBI
Human
Observer
Interface
FBP
Numerical Observer Results: RBI
Numerical Observer Results: FBP
Human Observer Results
Conclusions
• Iterative RBI-EM including all corrections
performs better than that of FBP.
• The best performance reconstruction
strategy is RBI-EM with 5 iteration and 1
pixel FWHM in Gaussian post-filtering.
• Numerical observer with and without mean
background subtraction set the upper and
lower bounds achievable by human
observer.
Acknowlegements
• This work was supported by a Charles
Sturt University Special Study grant and a
NIH research grant.
• The co-authors of this work are Prof Mike
King, Dr Howard Gifford and Dr Hennie
Pretorius at the University of
Massachusetts Medical School.