Kwok L. Lam

Measurement of prostate movement over the course of routine radiotherapy using implanted markers

PURPOSE To measure the range and frequency of occurrence of intertreatment movement of the prostate gland over the course of radiotherapy, and to demonstrate that the prostate may move independently of the surrounding bones of the pelvis. METHODS AND MATERIALS Ten patients underwent implantation of radiopaque markers around the prostate. Orthogonal portal films were taken at multiple stages during the course of treatment and digitized. An image registration tool was used to solve for film detector placement and, subsequently, to determine positional changes between structures on a reference portal image pair and all subsequent pairs for each patient. Transformations describing prostate movement were measured independently of those describing setup variations of the pelvic girdle. RESULTS Translation and/or rotation of the prostate was detected in 70% of the treatments for which films were taken. The maximum measured displacement was 7.5 mm along a major axis. Typical translations of the prostate were between 0-4 mm. The translation and rotation had a predominant direction, suggesting a natural axis for prostate movement. CONCLUSION Although significant prostate displacement can occur between treatments, the typical range of movement seen along a major axis was less than 5 mm. Proper treatment planning should consider the movement of the target independent of surrounding bony anatomy. Advances in online portal imaging, image registration, and dynamic field shaping may permit shaped fields that encompass the prostate gland in its position at the time of treatment, allowing for the use of smaller fields while ensuring proper target coverage.

Uncertainties in CT-based radiation therapy treatment planning associated with patient breathing

PURPOSE To evaluate uncertainties associated with treatment-planning computed tomography (CT) data obtained with the patient breathing freely. METHODS AND MATERIALS Patients with thoracic or abdominal tumors underwent a standard treatment-planning CT study while breathing quietly and freely, followed by CT scans while holding their breath at normal inhalation and normal exhalation. Identical treatment plans on all three CT data sets for each patient pointed out differences in: (a) radiation path lengths; (b) positions of the organs; (c) physical volumes of the lung, liver, and kidneys; (d) the interpretation of plan evaluation tools such as dose-volume histograms and normal tissue complication probability (NTCP) models; and (e) how well the planning CT data set represented the average of the inhalation and exhalation studies. RESULTS Inhalation and exhalation data differ in terms of radiation path length (nearly one quarter of the cases had path-length differences > 1 cm), although the free breathing and average path lengths do not exhibit large differences (0-9 mm). Liver and kidney movements averaged 2 cm, whereas differences between the free breathing and average positions averaged 0.6 cm. The physical volume of the liver between the free breathing and static studies varied by as much as 12%. The NTCP calculations on exhale and inhale studies varied from 3 to 43% for doses that resulted in a 15% NTCP on the free-breathing studies. CONCLUSION Free-breathing CT studies may improperly estimate the position and volume of critical structures, and thus may mislead evaluation of plans based on such volume dependent criteria such as dose-volume histograms and NTCP calculations.

Computerized analysis of mammographic microcalcifications in morphological and texture feature spaces

We are developing computerized feature extraction and classification methods to analyze malignant and benign microcalcifications on digitized mammograms. Morphological features that described the size, contrast, and shape of microcalcifications and their variations within a cluster were designed to characterize microcalcifications segmented from the mammographic background. Texture features were derived from the spatial gray-level dependence (SGLD) matrices constructed at multiple distances and directions from tissue regions containing microcalcifications. A genetic algorithm (GA) based feature selection technique was used to select the best feature subset from the multi-dimensional feature spaces. The GA-based method was compared to the commonly used feature selection method based on the stepwise linear discriminant analysis (LDA) procedure. Linear discriminant classifiers using the selected features as input predictor variables were formulated for the classification task. The discriminant scores output from the classifiers were analyzed by receiver operating characteristic (ROC) methodology and the classification accuracy was quantified by the area, Az, under the ROC curve. We analyzed a data set of 145 mammographic microcalcification clusters in this study. It was found that the feature subsets selected by the GA-based method are comparable to or slightly better than those selected by the stepwise LDA method. The texture features (Az = 0.84) were more effective than morphological features (Az = 0.79) in distinguishing malignant and benign microcalcifications. The highest classification accuracy (Az = 0.89) was obtained in the combined texture and morphological feature space. The improvement was statistically significant in comparison to classification in either the morphological (p = 0.002) or the texture (p = 0.04) feature space alone. The classifier using the best feature subset from the combined feature space and an appropriate decision threshold could correctly identify 35% of the benign clusters without missing a malignant cluster. When the average discriminant score from all views of the same cluster was used for classification, the Az value increased to 0.93 and the classifier could identify 50% of the benign clusters at 100% sensitivity for malignancy. Alternatively, if the minimum discriminant score from all views of the same cluster was used, the Az value would be 0.90 and a specificity of 32% would be obtained at 100% sensitivity. The results of this study indicate the potential of using combined morphological and texture features for computer-aided classification of microcalcifications.

Daily prostate targeting using implanted radiopaque markers

PURPOSE A system has been implemented for daily localization of the prostate through radiographic localization of implanted markers. This report summarizes an initial trial to establish the accuracy of patient setup via this system. METHODS AND MATERIALS Before radiotherapy, three radiopaque markers are implanted in the prostate periphery. Reference positions are established from CT data. Before treatment, orthogonal radiographs are acquired. Projected marker positions are extracted semiautomatically from the radiographs and aligned to the reference positions. Computer-controlled couch adjustment is performed, followed by acquisition of a second pair of radiographs to verify prostate position. Ten patients (6 prone, 4 supine) participated in a trial of daily positioning. RESULTS Three hundred seventy-four fractions were treated using this system. Treatment times were on the order of 30 minutes. Initial prostate position errors (sigma) ranged from 3.1 to 5.8 mm left-right, 4.0 to 10.1 mm anterior-posterior, and 2.6 to 9.0 mm inferior-superior in prone patients. Initial position was more reproducible in supine patients, with errors of 2.8 to 5.0 mm left-right, 1.9 to 3.0 mm anterior-posterior, and 2.6 to 5.3 mm inferior-superior. After prostate localization and adjustment, the position errors were reduced to 1.3 to 3.5 mm left-right, 1.7 to 4.2 mm anterior-posterior, and 1.6 to 4.0 mm inferior-superior in prone patients, and 1.2 to 1.8 mm left-right, 0.9 to 1.8 mm anterior-posterior, and 0.8 to 1.5 mm inferior-superior in supine patients. CONCLUSIONS Daily targeting of the prostate has been shown to be technically feasible. The implemented system provides the ability to significantly reduce treatment margins for most patients with cancer confined to the prostate. The differences in final position accuracy between prone and supine patients suggest variations in intratreatment prostate movement related to mechanisms of patient positioning.

Automated localization of the prostate at the time of treatment using implanted radiopaque markers: technical feasibility.

PURPOSE Prostate movement is a major consideration in the formation of target volumes for conformal radiation therapy of prostate cancer. The goal of this study was to determine the technical feasibility of using implanted radiopaque markers and digital imaging to localize the prostate at the time of treatment, thus allowing for reduction of the margin required for uncertainty in target position. METHODS AND MATERIALS Radiopaque markers implanted around the prostate prior to treatment are visible on electronic radiographs generated with a portal imager or diagnostic imaging device. The locations of the images of these markers on the digital radiographs were automatically determined by a template-matching algorithm. The coordinates of the markers were found by projecting rays through the marker locations on orthogonal radiographs using a three-dimensional (3D) point-matching algorithm. Prostate and/or patient movement was inferred from the marker displacements. Images generated from known movements of a phantom with implanted markers were tested with this algorithm. Locations of markers from daily images of patients with implanted markers were determined by both manual and automatic techniques to determine the efficacy of automated localization on typical clinical images. RESULTS Prostate movements can be automatically detected in a phantom using low-energy photons within 30 s after image acquisition and with a precision of better than 1 mm in translation and 1 degree in rotation (indistinguishable from the uncertainty in measuring precision). CONCLUSION The studies show that on-line repositioning of the patient based on localization of the markers at the time of treatment is feasible, and may reduce the uncertainty in prostate location when combined with practical on-line repositioning techniques.

Computer‐aided detection of mammographic microcalcifications: Pattern recognition with an artificial neural network

We are developing a computer program for automated detection of clustered microcalcifications on mammograms. In this study, we investigated the effectiveness of a signal classifier based on a convolution neural network (CNN) approach for improvement of the accuracy of the detection program. Fifty-two mammograms with clustered microcalcifications were selected from patient files. The clusters on the mammograms were ranked by experienced mammographers and divided into an obvious group, an average group, and a subtle group. The average and subtle groups were combined and randomly divided into two sets, each of which was used as training or test set alternately. The obvious group served as an additional independent test set. Regions of interest (ROIs) containing potential individual microcalcifications were first located on each mammogram by the automated detection program. The ROIs from one set of the mammograms were used to train CNNs of different configurations with a back-propagation method. The generalization capability of the trained CNNs was then examined by their accuracy of classifying the ROIs from the other set and from the obvious group. The classification accuracy of the CNNs for the ROIs was evaluated by receiver operating characteristic (ROC) analysis. It was found that CNNs of many different configurations can reach approximately the same performance level, with the area under the ROC curve (Az) of 0.9. We incorporated a trained CNN into the detection program and evaluated the improvement of the detection accuracy by the CNN using free response ROC analysis. Our results indicated that, over a wide range of true-positive (TP) cluster detection rate, the CNN classifier could reduce the number of false-positive (FP) clusters per image by more than 70%. For the obvious cases, at a TP rate of 100%, the FP rate reduced from 0.35 cluster per image to 0.1 cluster per image. For the average and subtle cases, the detection accuracy improved from a TP rate of 87% at an FP rate of four clusters per image to a TP rate of 90% at an FP rate of 1.5 clusters per image.

Computerized classification of malignant and benign microcalcifications on mammograms: texture analysis using an artificial neural network.

We investigated the feasibility of using texture features extracted from mammograms to predict whether the presence of microcalcifications is associated with malignant or benign pathology. Eighty-six mammograms from 54 cases (26 benign and 28 malignant) were used as case samples. All lesions had been recommended for surgical biopsy by specialists in breast imaging. A region of interest (ROI) containing the microcalcifications was first corrected for the low-frequency background density variation. Spatial grey level dependence (SGLD) matrices at ten different pixel distances in both the axial and diagonal directions were constructed from the background-corrected ROI. Thirteen texture measures were extracted from each SGLD matrix. Using a stepwise feature selection technique, which maximized the separation of the two class distributions, subsets of texture features were selected from the multi-dimensional feature space. A backpropagation artificial neural network (ANN) classifier was trained and tested with a leave-one-case-out method to recognize the malignant or benign microcalcification clusters. The performance of the ANN was analysed with receiver operating characteristic (ROC) methodology. It was found that a subset of six texture features provided the highest classification accuracy among the feature sets studied. The ANN classifier achieved an area under the ROC curve of 0.88. By setting an appropriate decision threshold, 11 of the 28 benign cases were correctly identified (39% specificity) without missing any malignant cases (100% sensitivity) for patients who had undergone biopsy. This preliminary result indicates that computerized texture analysis can extract mammographic information that is not apparent by visual inspection. The computer-extracted texture information may be used to assist in mammographic interpretation, with the potential to reduce biopsies of benign cases and improve the positive predictive value of mammography.

Relative dosimetry using active matrix flat-panel imager (AMFPI) technology

The first examination of the use of active matrix flat-panel arrays for dosimetry in radiotherapy is reported. Such arrays are under widespread development for diagnostic and radiotherapy imaging. In the current study, an array consisting of 512 x 512 pixels with a pixel pitch of 508 microm giving an area of 26 x 26 cm2 has been used. Each pixel consists of a light sensitive amorphous silicon (a-Si:H) photodiode coupled to an a-Si:H thin-film transistor. Data was obtained from the array using a dedicated electronics system allowing real-time data acquisition. In order to examine the potential of such arrays as quality assurance devices for radiotherapy beams, field profile data at photon energies of 6 and 15 MV were obtained as a function of field size and thickness of overlying absorbing material (solid water). Two detection configurations using the array were considered: a configuration (similar to the imaging configuration) in which an overlying phosphor screen is used to convert incident radiation to visible light photons which are detected by the photodiodes; and a configuration without the screen where radiation is directly sensed by the photodiodes. Compared to relative dosimetry data obtained with an ion chamber, data taken using the former configuration exhibited significant differences whereas data obtained using the latter configuration was generally found to be in close agreement. Basic signal properties, which are pertinent to dosimetry, have been investigated through measurements of individual pixel response for fluoroscopic and radiographic array operation. For signal levels acquired within the first 25% of pixel charge capacity, the degree of linear response with dose was found to be better than 99%. The independence of signal on dose rate was demonstrated by means of stability of pixel response over the range of dose rates allowed by the radiation source (80-400 MU/min). Finally, excellent long-term stability in pixel response, extending over a 2 month period, was observed.

Daily targeting of intrahepatic tumors for radiotherapy

INTRODUCTION A system has been developed for daily targeting of intrahepatic tumors using a combination of ventilatory immobilization, in-room diagnostic imaging, and on-line setup adjustment. By reducing geometric position uncertainty, as well as organ movement, this system permits reduction of margins and thus potentially higher treatment doses. This paper reports our initial experience treating 8 patients with focal liver tumors using this system. METHODS AND MATERIALS The system includes diagnostic X-ray tubes mounted on the wall and ceiling of a treatment room, an active matrix flat panel imager, in-room control for image acquisition and setup adjustment, and a ventilatory immobilization system via active breathing control (ABC). Eight patients participated in the study, two using an early prototype ABC unit, and the remaining six with a commercial ABC system and improved setup measurement tools. Treatment margins were reduced, and dose consequently increased because of increased confidence in target position under this protocol. After daily setup via skin marks, orthogonal radiographs were acquired at suspended ventilation. The images were aligned to the CT model using the diaphragm for inferior-superior (IS) alignment, and the skeleton for left-right (LR) and anterior-posterior (AP) alignment. Adjustments were made for positioning errors greater than a threshold (3 or 5 mm). After treatment, retrospective analysis determined the final setup accuracy, as well as the error in initial setup measurement performed before setup adjustment. RESULTS Two hundred sixty-two treatment fractions were delivered on eight patients, with 171 treatments requiring repositioning. Typical treatment times were 25-30 min. Patients were able to tolerate ABC throughout the course of treatment. Breath holds up to 35 s long were used for treatment. The use of on-line imaging and setup adjustment reduced setup errors (sigma) from 4.0 mm (LR), 6.7 mm (IS), and 3.8 mm (AP) to 2.1 mm (LR), 3.5 mm (IS), and 2.3 mm (AP). Prescribed doses were increased using this system by an average of 5 Gy. CONCLUSIONS Daily targeting of intrahepatic targets has been demonstrated to be feasible. The potential for reduction in treatment margin and consequential safe dose escalation has been demonstrated, while maintaining reasonable treatment delivery times.

TG-69 : Radiographic film for megavoltage beam dosimetry

TG-69 is a task group report of the AAPM on the use of radiographic film for dosimetry. Radiographic films have been used for radiation dosimetry since the discovery of x-rays and have become an integral part of dose verification for both routine quality assurance and for complex treatments such as soft wedges (dynamic and virtual), intensity modulated radiation therapy (IMRT), image guided radiation therapy (IGRT), and small field dosimetry like stereotactic radiosurgery. Film is convenient to use, spatially accurate, and provides a permanent record of the integrated two dimensional dose distributions. However, there are several challenges to obtaining high quality dosimetric results with film, namely, the dependence of optical density on photon energy, field size, depth, film batch sensitivity differences, film orientation, processing conditions, and scanner performance. Prior to the clinical implementation of a film dosimetry program, the film, processor, and scanner need to be tested to characterize them with respect to these variables. Also, the physicist must understand the basic characteristics of all components of film dosimetry systems. The primary mission of this task group report is to provide guidelines for film selection, irradiation, processing, scanning, and interpretation to allow the physicist to accurately and precisely measure dose with film. Additionally, we present the basic principles and characteristics of film, processors, and scanners. Procedural recommendations are made for each of the steps required for film dosimetry and guidance is given regarding expected levels of accuracy. Finally, some clinical applications of film dosimetry are discussed.