Regularization design for breast lesion detection in penalized maximum likelihood image reconstruction

Published on May 2, 2012 in ISBI (International Symposium on Biomedical Imaging)
· DOI :10.1109/ISBI.2012.6235626
Li Yang4
Estimated H-index: 4
(UC Davis: University of California, Davis),
Jian Zhou140
Estimated H-index: 140
(UC Davis: University of California, Davis),
Jinyi Qi52
Estimated H-index: 52
(UC Davis: University of California, Davis)
Sources
Abstract
Detecting cancerous lesion is a major clinical application in emission tomography. In a previous work, we have shown that penalized maximum likelihood image reconstruction can improve lesion detection at a fixed location by designing a shift-invariant quadratic penalty function. Here we extend this work to detection of tumors at unknown positions. We present a method to design a shift-variant quadratic penalty function that maximizes the detectability of lesions at all possible locations. We conducted computer-based Monte Carlo simulations to compare the optimized shift-variant penalty with the conventional penalty for detecting a breast lesion. Lesion detectability was assessed by a channelized Hotelling observer and human observer. The results showed a statistically significant improvement in lesion detection by using the optimized shift-variant penalty function compared to using the conventional penalty function.
📖 Papers frequently viewed together
30 Citations
4 Citations
2004
1 Author (Jinyi Qi)
3 Citations
References7
Newest
#1Jinyi Qi (UC Davis: University of California, Davis)H-Index: 52
#2Ronald H. Huesman (LBNL: Lawrence Berkeley National Laboratory)H-Index: 24
Detecting cancerous lesions is one major application in emission tomography. In this paper, we study penalized maximum-likelihood image reconstruction for this important clinical task. Compared to analytical reconstruction methods, statistical approaches can improve the image quality by accurately modelling the photon detection process and measurement noise in imaging systems. To explore the full potential of penalized maximum-likelihood image reconstruction for lesion detection, we derived simp...
30 CitationsSource
#1Jinyi Qi (University of California, Berkeley)H-Index: 52
Statistical image reconstruction methods based on maximum a posteriori (MAP) principle have been developed for emission tomography. The prior distribution of the unknown image plays an important role in MAP reconstruction. The most commonly used prior is the Gaussian prior, whose logarithm has a quadratic form. Gaussian priors are relatively easy to analyze. It has been shown that the effect of a Gaussian prior can be approximated by linear-filtering a maximum likelihood (ML) reconstruction. As ...
9 CitationsSource
#1Jinyi Qi (LBNL: Lawrence Berkeley National Laboratory)H-Index: 52
Bayesian methods based on the maximum a posteriori principle (also called penalized maximum-likelihood methods) have been developed to improve image quality in emission tomography. To explore the full potential of Bayesian reconstruction for lesion detection, we derive simplified theoretical expressions that allow fast evaluation of the detectability of a lesion in Bayesian reconstruction. This work is built on the recent progress on the theoretical analysis of image properties of statistical re...
47 CitationsSource
#1J.W. Stayman (UM: University of Michigan)H-Index: 5
#2Jeffrey A. Fessler (UM: University of Michigan)H-Index: 78
Traditional space-invariant regularization methods in tomographic image reconstruction using penalized-likelihood estimators produce images with nonuniform spatial resolution properties. The local point spread functions that quantify the smoothing properties of such estimators are space variant, asymmetric, and object-dependent even for space invariant imaging systems. The authors propose a new quadratic regularization scheme for tomographic imaging systems that yields increased spatial uniformi...
180 CitationsSource
#1Jinyi Qi (SC: University of Southern California)H-Index: 52
#2Richard M. Leahy (SC: University of Southern California)H-Index: 74
The authors examine the spatial resolution and variance properties of PET images reconstructed using maximum a posteriori (MAP) or penalized-likelihood methods. Resolution is characterized by the contrast recovery coefficient (CRC) of the local impulse response. Simplified approximate expressions are derived for the local impulse response CRCs and variances for each voxel. Using these results the authors propose a practical scheme for selecting spatially variant smoothing parameters to optimize ...
154 CitationsSource
#1Arthur E. Burgess (RIT: Rochester Institute of Technology)H-Index: 3
The receiver operating characteristic (ROC) method has been successfully used in medical imaging for 20 years. It has been so successful that many people think of it as an end in itself rather than just one of several ways to assess human observer performance of image based decision tasks. Studies of human and ideal observer decision performance are designed to estimate a measure of observer performance (e.g., efficiency, d′ or d a ). The experimenter would like to obtain accurate and precise es...
170 CitationsSource
#1Jie Yao (UA: University of Arizona)H-Index: 3
#2Harrison H. Barrett (UA: University of Arizona)H-Index: 64
A psychophysical experiment was conducted to measure the performance of human observers in detecting an exactly known signal against a random, nonuniform background in the presence of noise correlations introduced by post-detection filtering (postprocessing). In order to predict this human performance, a new model observer was synthesized by adding frequency-selective channels to the Hotelling observer model which we have previously used for assessment of image quality. This new `channelized' Ho...
115 CitationsSource
Cited By5
Newest
#1Lishui Cheng (JHUSOM: Johns Hopkins University School of Medicine)H-Index: 9
#2Robert F. Hobbs (JHUSOM: Johns Hopkins University School of Medicine)H-Index: 59
Last. Eric C. Frey (JHUSOM: Johns Hopkins University School of Medicine)H-Index: 51
view all 4 authors...
Purpose: Three-dimensional (3D) dosimetry has the potential to provide better prediction of response of normal tissues and tumors and is based on 3D estimates of the activity distribution in the patient obtained from emission tomography. Dose–volume histograms (DVHs) are an important summary measure of 3D dosimetry and a widely used tool for treatment planning in radiation therapy. Accurate estimates of the radioactivity distribution in space and time are desirable for accurate 3D dosimetry. The...
2 CitationsSource
#1Li Yang (UC Davis: University of California, Davis)H-Index: 4
#2Andrea Ferrero (UC Davis: University of California, Davis)H-Index: 8
Last. Jinyi Qi (UC Davis: University of California, Davis)H-Index: 52
view all 5 authors...
Detecting cancerous lesions is a major clinical application in emission tomography. Previously, we developed a method to design a shift-variant quadratic penalty function in penalized maximum-likelihood (PML) image reconstruction to improve the lesion detectability. We used a multiview channelized Hotelling observer (mvCHO) to assess the lesion detectability in three-dimensional images and validated the penalty design using computer simulations. In this study, we evaluate the benefit of the prop...
4 CitationsSource
#1Li Yang (UC Davis: University of California, Davis)H-Index: 4
#2Jian Zhou (UC Davis: University of California, Davis)H-Index: 12
Last. Jinyi Qi (UC Davis: University of California, Davis)H-Index: 52
view all 5 authors...
Detecting cancerous lesions is a major clinical application in emission tomography. In previous work, we have studied penalized maximum-likelihood (PML) image reconstruction for the detection task and proposed a method to design a shift-invariant quadratic penalty function to maximize detectability of a lesion at a known location in a two dimensional image. Here we extend the regularization design to maximize detectability of lesions at unknown locations in fully 3D PET. We used a multiview chan...
24 CitationsSource
#1Joyita Dutta (Harvard University)H-Index: 13
#2Sangtae Ahn (Harvard University)H-Index: 21
Last. Quanzheng Li (Harvard University)H-Index: 32
view all 3 authors...
Quantitative measures of image quality and reliability are critical for both qualitative interpretation and quantitative analysis of medical images. While, in theory, it is possible to analyze reconstructed images by means of Monte Carlo simulations using a large number of noise realizations, the associated computational burden makes this approach impractical. Additionally, this approach is less meaningful in clinical scenarios, where multiple noise realizations are generally unavailable. The pr...
41 CitationsSource
#1Sangtae AhnH-Index: 21
#2Quanzheng LiH-Index: 32
Quantitative measures of image quality and reliability are critical for both qualitative interpretation and quantitative analysis of medical images. While, in theory, it is possible to analyze reconstructed images by means of Monte Carlo simulations using a large number of noise realizations, the associated computational burden makes this approach impractical. Additionally, this approach is less meaningful in clinical scenarios, where multiple noise realizations are generally unavailable. The pr...