Positron Emission Mammography: Initial Clinical Results

of 6

Please download to get full document.

View again

All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
PDF
6 pages
0 downs
0 views
Share
Description
Background: Evaluation of high-risk mammograms represents an enormous clinical challenge. Functional breast imaging coupled with mammography (positron emission mammography [PEM]) could improve imaging of such lesions. A prospective study was
Tags
Transcript
  Positron Emission Mammography: Initial Clinical Results Edward A. Levine, MD, Rita I. Freimanis, MD, Nancy D. Perrier, MD, Kathryn Morton, MD,Nadia M. Lesko, MD, Simon Bergman, MD, Kim R. Geisinger, MD, Rodney C. Williams, MS,Connie Sharpe, MBA, Valera Zavarzin, MS, Irving N. Weinberg, MD, PhD,Pavel Y. Stepanov, MS, David Beylin, MS, Kathryn Lauckner, PhD, Mohan Doss, PhD,Judy Lovelace, RN, and Lee P. Adler, MD Background:  Evaluation of high-risk mammograms represents an enormous clinical challenge.Functional breast imaging coupled with mammography (positron emission mammography [PEM])could improve imaging of such lesions. A prospective study was performed using PEM in womenscheduled for stereotactic breast biopsy. Methods:  Patients were recruited from the surgical clinic. Patients were injected with 10 mCi of 2-[ 18 F] fluorodeoxyglucose. One hour later, patients were positioned on the stereotactic biopsytable, imaged with a PEM scanner, and a stereotactic biopsy was performed. Imaging was reviewedand compared with pathologic results. Results:  There were 18 lesions in 16 patients. PEM images were analyzed by drawing a regionof interest at the biopsy site and comparing the count density in the region of interest with thebackground. A lesion-to-background ratio  2.5 appeared to be a robust indicator of malignancy andyielded a sensitivity of 86%, specificity of 91%, and overall diagnostic accuracy of 89%. No adverseevents were associated with the PEM imaging. Conclusions:  The data show that PEM is safe, feasible, and has an encouraging accuracy rate inthis initial experience. Lesion-to-background ratios  2.5 were found to be a useful threshold valuefor identifying positive (malignant) results. This study supports the further development of PEM. Key Words:  Mammography—Breast—Imaging—PET—Cancer—Diagnosis. One of the major goals of breast imaging is to detectsmall breast cancers so that treatment may be initiatedwhen the outlook is most favorable. Once a cancer hasbeen detected, it is useful to demonstrate the extent of thelesion for surgical planning and to identify other foci of tumor in the breast. These tasks are not reliably per-formed even with the best tools currently in general use(mammography and ultrasound). 1 A key reason for thelimited efficacy of conventional imaging is the limitationto anatomic depictions. With the addition of functionalimaging, however, it is possible that mammographicallyoccult disease may be better characterized, benign le-sions may be recognized as such, and unsuspected tu-mors may be detected.Positron emission tomography (PET) allows the visu-alization of functional metabolism in vivo. However,existing commercial PET instruments are not optimal forimaging small breast cancers. PET instruments custom-ized for breast applications have been shown to exhibitsuperior technical characteristics for imaging small le-sions in breast phantoms. 2 With the goal of evaluating adedicated prototype instrument for breast PET, a pro-spective pilot study was designed to develop methodol-ogy. We used this instrument (positron emission mam-mography [PEM] scanner) mounted on a stereotactic Received March 15, 2002; accepted August 5, 2002.From the Departments of Surgery (EAL, NDP, JL), Radiology (RIF,KM, NML, CS, RCW), and Pathology (SB, KRG), Wake ForestUniversity School of Medicine, Winston-Salem, North Carolina; PEMTechnologies Inc. (VZ, INW, PYS, DB), Bethesda, Maryland; Seleongmbh (KL), Freiburg, Germany; and Department of Nuclear Medicine(MD, LPA), Fox Chase Cancer Center, Philadelphia, Pennsylvania.Presented at the Society of Surgical Oncology meeting, Denver, CO,March 15–17, 2002.Address correspondence and reprint requests to: Edward A. Levine,MD, Surgical Oncology Service, Wake Forest University, MedicalCenter Blvd., Winston-Salem, NC 27157; Fax : 336-716-9758; E-mail:elevine@wfubmc.edu. Published by Lippincott Williams & Wilkins © 2003 The Society of SurgicalOncology, Inc.  Annals of Surgical Oncology,  10 (1):86–91 DOI: 10.1245/ASO.2003.03.047 86   biopsy table. 3 To facilitate correlation between PEMfindings and pathology results, it was necessary to devisea scanning method that would facilitate both imaging of,and access to, the biopsy site. This article describes ourinitial experience with this novel imaging modality. METHODSClinical Protocol Human subjects were enrolled in an Institutional Re-view Board approved clinical protocol at Wake ForestUniversity-Baptist Medical Center. All eligible subjectswere seen in the breast surgery clinic for suspiciousmammogram(s), with or without suspicious physicalfindings requiring biopsy. Of those patients entered intothe protocol, the first set of subjects to have completedata and follow-up (18 breast lesions in 17 breasts from16 subjects) were selected for analysis.Following 4 hours of fasting, a sample of blood fromeach subject was obtained for serum glucose determina-tion. The patient received an intravenous injection of 10mCi of   18 F-fluorodeoxyglucose (FDG). One hour later,the patient was placed prone on the stereotactic biopsytable (Lorad, Danbury, CT). The breast was positionedso that the suspicious lesion was in the field of view of the biopsy window.The prototype PEM scanner consists of detector headsplaced on each side of the compressed breast and asso-ciated acquisition and display electronics. One mountingframe is attached to the compression paddle and the otherto the back of the detector plate of the prone x-ray tableto allow rapid insertion and removal of the components(Fig. 1). 1 This prototype device has a field of view of 5.6by 5.6 cm, similar to the images obtained on stereotacticprone x-ray tables equipped with digital spot mammog-raphy. Spatial resolution is better than 3-mm full-widthat half-maximum, 2 significantly better than can beachieved with conventional whole-body PET scanners.The PEM detector heads can be removed from the x-raytable and replaced within seconds, without requiringrelease of the subject’s breast from compression.The following imaging procedure was performed foreach subject on the stereotactic biopsy table. A floodfield calibration examination of the PEM device wasobtained using a flat phantom filled with  18 FDG (100mCi) placed midway between the PEM detector heads.The flood field was imaged between 4 and 10 minutes.The PEM camera heads were then attached to the com-pression paddle and back plate (Fig. 1). A 120-secondposition calibration check of the PEM device was per-formed with a cardboard insert containing two Na-22positron-emitting point sources. The PEM camera headswere removed and an x-ray image of the two pointsources was taken. This position calibration was appliedto insure detector efficiency and x-ray/PEM alignment.The PEM camera heads are then reattached to the com-pression paddles. The patient was then placed in theprone position on the table, resting on a 2-mm thick layerof sheet lead placed under the table’s cushion. The sub- ject breast was then inserted into the table aperture, anda scout x-ray view of the subject’s breast was takenconfirming that the mammographic abnormality is visi-ble in the field-of-view. Without releasing breast com-pression, the PEM detector heads are then reintroducedon both sides of the compressed breast, and a 4-minuteexamination of the suspicious area is obtained. The PEMdetector heads were then removed from the table, and thesubject underwent biopsy of the mammographically sus-picious lesion. Routine stereo, prefire, and postfire viewswere obtained. Using standard sterile procedure, breastbiopsy was performed with an 11-gauge vacuum-assistedcore biopsy needle (Mammotome, Ethicon Endo-Sur-gery, Cincinnati OH). Cores of breast tissue were takensequentially at 12 locations around the clock at 2-hourintervals (e.g., 2:00, 4:00, 6:00, 8:00, 10:00, 12:00, 1:00,3:00, 5:00, 7:00, 9:00, 11:00). Each core was examinedby a pathologist without knowledge of the PEM studyresults. For those subjects who were biopsied on thetable, the coordinates of the lesion location were re-corded for later analysis. In one case, a subject with anobvious cancer was not biopsied on the table on the dayof the PEM examination. Subjects did not report unto-ward effects as a result of the examinations with the PEMdevice. FIG. 1.  PEM detectors mounted on a sterotactic breast biopsy sys-tem. Illustrated are: 1 (compression paddle), 2 (back plate), and 3 and4 (detectors). 87 POSITRON EMISSION MAMMOGRAPHY   Ann Surg Oncol, Vol. 10, No. 1, 2003  Data Acquisition Protocol The PEM device recorded a file of gamma ray eventsacquired by the PEM detector heads. This list-mode fileincluded coincident events, as well as delayed coincidentevents (i.e., randoms) for postprocessing randoms cor-rection. Back-projected images taken with the PEM cam-era were visible within 2 minutes of acquisition. X-rayimages were ported from the stereotactic biopsy com-puter to the PEM computer using a file transfer protocol-like protocol implemented with data transfer from thebiopsy table computer parallel port, as describedpreviously. 2 Data Analysis The list-mode data acquisition files were reconstructedwith an iterative maximum likelihood estimation methodusing median filtering and 2-mm in-plane pixel size. Indeference to the standard system for spatial localizationon the biopsy unit, the planes were assigned x- andy-axes, and the vector between the compression paddleswas assigned to the z direction. Twenty-four equidistantz planes of reconstruction were selected that were paral-lel to the compression paddles. Due to variable breastthickness as indicated by the width of compression onthe biopsy table, the z width of each plane varies inthickness from approximately 2 to 3.5 mm, depending oncompression distance. Visualization is accomplished us-ing IDL software (IDL version 5.4, Research Systems,Boulder, CO). Using the point source images for bothPEM and x-ray to correct for x-y shift and magnification,the two image modalities were co-registered.For those subjects undergoing stereotactic biopsy (17breasts/18 lesions), the target x, y, and z coordinates of the biopsy site of interest were recorded by the biopsycomputer. These biopsy locations were displayed on thePEM slices and on the x-ray image. In the one subjectwho did not have a stereotactic biopsy, a biopsy locationwas drawn on the co-registered PEM and x-ray images inthe approximate center of an obvious lesion that wasdiagnosed by fine-needle aspiration then removed with amastectomy. Image Analysis Regions of interest (ROI) were created on the PEMimages (Fig. 2) at two locations. The first location cor-responds to the site of stereotactic core biopsy and thesecond to a representative background region of compa-rable size and tissue thickness. The biopsy site ROI wascentered at the biopsy location specified by x, y, and zcoordinates for the biopsy procedure. The most clinicallysignificant (malignant) histology in the biopsy cores isrecorded as the “gold standard” result, and the PEMvalue recorded for that tissue is the highest pixel inten-sity in the volume of tissue biopsied. This three-dimen-sional biopsy ROI measurement was designed to modelthe sample geometry of the core biopsy. The sampledvolume maximally includes tissue approximately 1 cm indiameter around the biopsy core needle in the x-y planeand approximately 19 mm along the biopsy z-axis, cen-tered around the z coordinate. To mathematically modelthe geometry of the biopsy sampling for histopathology,the measurement of the value of the biopsy ROI wascomputed as the maximum value of the five slices sur-rounding the biopsy z value, where each slice representsapproximately 2 mm thickness.For the background ROI determination, the z sliceselected corresponded to a depth comparable to thatdetermined by the biopsy target slice (z location). Twoboard-certified radiologists collaborated in identifying arepresentative background region in that slice whichcorresponded to an area of low attenuation (i.e., fattybreast tissue) as identified by the digital x-ray of thebreast, at a site of comparable thickness to the site of biopsy. The mean and minimum pixel count intensitieswere calculated. There are several possible methods of selecting the appropriate value for the background ROI.To model the physiological concept of lesion-to-back-ground ratio with the highest fidelity, the minimum pixelvalue is used because this would represent the lowestdegree of malignancy in the ROI. The mean value for thebackground should provide better counting statistics thandoes the minimum value. As PET images typically havefew counts per pixel in the breast, the mean value haspreviously been used in the PET literature. 4 However themean value determination may vary more with userselection of the region of interest than would the mini-mum value. The count statistics advantage are likely tobe less important in the PEM images than with conven-tional PET scanners because count efficiency for activityin the breast are much higher in the PEM scanner than in FIG. 2.  Representative clinical images (subject 6b in Table 1). Im-ages from a patient study illustrating (from left to right) digital x-ray(with location of biopsy and background ROIs), reconstructed PEMimage from central plane of biopsy, and overlaid PEM/x-ray image.The lesion was invasive ductal carcinoma by histology. True positive. 88 E. A. LEVINE ET AL.  Ann Surg Oncol, Vol. 10, No. 1, 2003  a whole-body PET scanner. 2 For these reasons, resultsfrom both methods (mean and minimum) are reported.PEM lesion-to-background ratios are calculated forbiopsy location using both mean and minimum back-ground values. These ratios were compared with thehistological findings. The left-uppermost corner of thereceiver operating characteristic (ROC) curves was usedto identify the cutoff threshold ROI ratio for optimalsensitivity and specificity in identifying the presence orabsence of malignancy. RESULTS The PEM lesion-to-background ratios for the 18 le-sions in 17 breasts (16 patients) studied are listed inTable 1. The average age of the 16 women was 57 years,with a range of 34 to 84 years. Of the 16 study subjects,4 were black and 12 were white. No adverse reactions toFDG or the PEM imaging procedures were encountered.Among the 18 breast lesions evaluated, there were a totalof 7 with carcinoma (5 ductal and 1 each of the lobularand mucinous types) and 11 with benign findings. Thelesion-to-background ratios for the PEM studies variedbetween 1.32 and 7.7. Representative images are shownin Figures 1–4. The average ratio for the malignantlesions was 3.95 vs. 1.94 for the benign lesions,  P   .0032 (by Wilcoxon rank sum test).An ROC curve was generated for the lesion-to-back-ground data (using mean pixel density to determine thebackground value). Selecting the left-most and upper-most corner of the ROC curve, a cutoff lesion-to-back-ground ratio threshold of   2.5 was selected to segmentbenign from malignant lesions. With this cutoff, therewas one false positive and one false negative, with asensitivity in the detection of malignancy of 86%, spec-ificity of 91%, and overall diagnostic accuracy of 89%. TABLE 1.  Patient data Patientno.Pathologic diagnosis (most malignantportion of pathology specimen listed)Mean PEM lesion-to-background ratio ResultInvasive tumor sizeat pathology (if malignant)1 Benign 1.95 TN2 Fibroadenoma 1.59 TN3 Ductal hyperplasia 2.19 TN4 Apocrine metaplasia 1.30 TN5 Typical hyperplasia 2.78 FP6a a Ductal hyperplasia 1.32 TN6b a Invasive ductal carcinoma 7.70 TP 30 mm7 Mild ductal hyperplasia 2.27 TN8 Atypical lobular hyperplasia 1.48 TN9 Sclerosing adenosis 2.03 TN10 Fibroadenoma 1.38 TN11 Fibroadenoma, sclerosing adenosis 2.53 TN12 Ductal carcinoma in situ, microinvasion 3.22 TP 1 mm13 b Invasive lobular carcinoma 1.97 FN 10 mm14 Invasive ductal carcinoma 3.50 TP 7 mm15 Ductal carcinoma in situ, microinvasion 3.33 TP 1 mm16 Invasive ductal carcinoma 2.63 TP 11 mm17 Invasive mucinous carcinoma 5.29 TP 10 mmTN, true negative; TP, true positive; FN, false negative; FP, false positive; PEM, positron emission mammography. a Cases 6a and 6b are from different breasts in the same patient. b Case 13 involved a patient with a lobular cancer who had discontinued hormone replacement therapy 1 week before examination with theprototype PEM scanner. FIG. 3.  Subject 11. Fibroadenoma and sclerosing adenosis. Ratio2.53. True negative. 89POSITRON EMISSION MAMMOGRAPHY   Ann Surg Oncol, Vol. 10, No. 1, 2003  Using the minimum background pixel value to calculatelesion-to-background ratios, but with a different cutoff lesion-to-background ratio threshold (i.e., 2.95), resultedin the same accuracy. Figures 3 and 4 are exams selectedas representative of typical results. DISCUSSION Previous FDG-PET scan studies of patients with sus-pected breast cancer have reported relatively low sensi-tivity in the detection of primary breast cancer. 4,5 Thesensitivity of FDG-PET is lowest for small lesions. 5 Byusing high-resolution components to improve spatial res-olution and reduce inter-detector distance, count ratestatistics are improved. The goal was to improve diag-nostic accuracy in the detection and characterization of primary carcinoma in the breast. The results reportedhere are preliminary, obtained with a prototype PEMscanner. The results reported here are semi-quantitative.Even higher accuracy in characterizing mammographicfindings may be obtainable with PEM scanners if quan-titative analysis is achieved. To correlate x-ray (mam-mographic) and functional PEM data, it is critical that theimages be co-registered accurately. This cannot be ac-complished with conventional PET, and successful PEMdevices must be adapted to mammographic units to per-mit simultaneous or sequential analysis without releasingcompression.For the single false negative in the series, (case 13 inTable 1) it is interesting to note that she had an invasivelobular lesion. The lobular histology has been reported tobe associated with false-negative FDG-PET. 5 Further,the subject had discontinued estrogen therapy approxi-mately 1 week before the PEM study. This may havealtered the results, although the effect of hormonal re-placement on the results of the PEM scans will not beknown without additional research. It is also possible thatthe low-lesion uptake reflects low-angiogenic activity, ashas been described in the magnetic resonance imagingliterature to explain the lowered sensitivity of magneticresonance imaging for some breast cancers. 6 The devel-opment of larger field-of-view detectors or absolutequantitative indices of uptake could overcome thislimitation.The marriage of functional and conventional imaginghas substantial potential impact. In addition to evaluationfor regional 7,8 and distant disease, 9 FDG-PET may giveinsight into tumor response to systemic therapies. 10–12 The PEM device dovetails nicely with conventionalmammography for analysis of suspicious mammo-graphic lesions. In addition to finding early malignancy,such analysis could serve as a guide to the surgeon inplanning breast-conserving surgery. Further, the PEMdevice may be able to assess changes in proliferativerates of ductal cells, potentially serving as a surrogatemarker for hyperplasia.This pilot study with a prototype PEM scanner sug-gests that quantitative comparisons of lesion-to-back-ground FDG concentration result in promising sensitiv-ity, specificity, and accuracy in the characterization of breast lesions. Further work is required to implementabsolute quantification with scatter correction, to de-velop larger field-of-view detectors, and to systemati-cally identify patient factors that could enhance or com-plicate our ability to detect breast cancer with PEM.Additional studies seem warranted and are required todetermine whether this promising performance is borneout with larger cohorts of patients. REFERENCES 1. Kerilikowske K, Grady D, Barclay J, et al. Variability and accu-racy in mammographic interpretation using the American Collegeof Radiology Breast Imaging and Reporting System.  J Natl Cancer  Inst   1998;90:1801–9.2. Weinberg IN, Majewski S, Weisenberger AG, et al. Preliminaryresults for positron emission mammography: real-time functionalbreast imaging in a conventional mammography gantry.  Eur J Nucl Med   1996;23:804–6.3. Weinberg IN, Zawarzin V, Pani R, et al. Implementing PET-guided biopsy. Integrating functional imaging data with digitalx-ray mammography cameras. In: Mun SK, ed.  Medical Imaging FIG. 4.  Subject 6b. Invasive ductal carcinoma. Ratio 7.7. Truepositive. 90 E. A. LEVINE ET AL.  Ann Surg Oncol, Vol. 10, No. 1, 2003
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks