Gregory H. Chu

Computer-aided quantitative bone scan assessment of prostate cancer treatment response.

ObjectiveThe development and evaluation of a computer-aided bone scan analysis technique to quantify changes in tumor burden and assess treatment effects in prostate cancer clinical trials. MethodsWe have developed and report on a commercial fully automated computer-aided detection (CAD) system. Using this system, scan images were intensity normalized, and then lesions were identified and segmented by anatomic region-specific intensity thresholding. Detected lesions were compared against expert markings to assess the accuracy of the CAD system. The metrics Bone Scan Lesion Area, Bone Scan Lesion Intensity, and Bone Scan Lesion Count were calculated from identified lesions, and their utility in assessing treatment effects was evaluated by analyzing before and after scans from metastatic castration-resistant prostate cancer patients: 10 treated and 10 untreated. In this study, patients were treated with cabozantinib, a MET/vascular endothelial growth factor inhibitor resulting in high rates of resolution of bone scan abnormalities. ResultsOur automated CAD system identified bone lesion pixels with 94% sensitivity, 89% specificity, and 89% accuracy. Significant differences in changes from baseline were found between treated and untreated groups in all assessed measurements derived by our system. The most significant measure, Bone Scan Lesion Area, showed a median (interquartile range) change from baseline at week 6 of 7.13% (27.61) in the untreated group compared with −73.76% (45.38) in the cabozantinib-treated group (P=0.0003). ConclusionOur system accurately and objectively identified and quantified metastases in bone scans, allowing for interpatient and intrapatient comparison. It demonstrates potential as an objective measurement of treatment effects, laying the foundation for validation against other clinically relevant outcome measures.

Preliminary results of automated removal of degenerative joint disease in bone scan lesion segmentation

Whole-body bone scintigraphy (or bone scan) is a highly sensitive method for visualizing bone metastases and is the accepted standard imaging modality for detection of metastases and assessment of treatment outcomes. The development of a quantitative biomarker using computer-aided detection on bone scans for treatment response assessment may have a significant impact on the evaluation of novel oncologic drugs directed at bone metastases. One of the challenges to lesion segmentation on bone scans is the non-specificity of the radiotracer, manifesting as high activity related to non-malignant processes like degenerative joint disease, sinuses, kidneys, thyroid and bladder. In this paper, we developed an automated bone scan lesion segmentation method that implements intensity normalization, a two-threshold model, and automated detection and removal of areas consistent with non-malignant processes from the segmentation. The two-threshold model serves to account for outlier bone scans with elevated and diffuse intensity distributions. Parameters to remove degenerative joint disease were trained using a multi-start Nelder-Mead simplex optimization scheme. The segmentation reference standard was constructed manually by a panel of physicians. We compared the performance of the proposed method against a previously published method. The results of a two-fold cross validation show that the overlap ratio improved in 67.0% of scans, with an average improvement of 5.1% points.

Automated segmentation of tumors on bone scans using anatomy-specific thresholding

Quantification of overall tumor area on bone scans may be a potential biomarker for treatment response assessment and has, to date, not been investigated. Segmentation of bone metastases on bone scans is a fundamental step for this response marker. In this paper, we propose a fully automated computerized method for the segmentation of bone metastases on bone scans, taking into account characteristics of different anatomic regions. A scan is first segmented into anatomic regions via an atlas-based segmentation procedure, which involves non-rigidly registering a labeled atlas scan to the patient scan. Next, an intensity normalization method is applied to account for varying levels of radiotracer dosing levels and scan timing. Lastly, lesions are segmented via anatomic regionspecific intensity thresholding. Thresholds are chosen by receiver operating characteristic (ROC) curve analysis against manual contouring by board certified nuclear medicine physicians. A leave-one-out cross validation of our method on a set of 39 bone scans with metastases marked by 2 board-certified nuclear medicine physicians yielded a median sensitivity of 95.5%, and specificity of 93.9%. Our method was compared with a global intensity thresholding method. The results show a comparable sensitivity and significantly improved overall specificity, with a p-value of 0.0069.

Quantitative bone scan lesion area as an early surrogate outcome measure indicative of overall survival in metastatic castration-resistant prostate cancer.

179 Background: Bone Scan Lesion Area (BSLA) is a biomarker that can be computed semi-automatically from whole-body scintigraphic imaging as a measure of overall bone tumor burden. Initial development and validation, including correlation with outcomes, was performed in trial cohorts from a single drug treatment with controls in subjects with metastatic castrate-resistant prostate cancer (CRPC). A 30% increase/decrease in BSLA was defined as progression/response on bone scan. We hypothesize that, when applied to an independent treatment trial cohort with a different mechanism of drug action, baseline BSLA and Week 12 change post-treatment are predictive of a subjects overall survival. Methods: From an anonymized imaging research database a cohort of 198 CRPC subjects was identified who enrolled in a treatment trial (127 treated, 71 placebo). This cohort was independent of those used for biomarker development and initial validation, and involved a different mechanism of drug action. Subjects underwent sta...

Bone Tumor Segmentation on Bone Scans Using Context Information and Random Forests

Bone tumor segmentation on bone scans has recently been adopted as a basis for objective tumor assessment in several phase II and III clinical drug trials. Interpretation can be difficult due to the highly sensitive but non-specific nature of bone tumor appearance on bone scans. In this paper we present a machine learning approach to segmenting tumors on bone scans, using intensity and context features aimed at addressing areas prone to false positives. We computed the context features using landmark points, identified by a modified active shape model. We trained a random forest classifier on 100 and evaluated on 73 prostate cancer subjects from a multi-center clinical trial. A reference segmentation was provided by a board certified radiologist. We evaluated our learning based method using the Jaccard index and compared against the state of the art, rule based method. Results showed an improvement from 0.50 +/- 0.31 to 0.57 +/- 0.27. We found that the context features played a significant role in the random forest classifier, helping to correctly classify regions prone to false positives.

Abstract B114: Cabozantinib effects on bone metastasis: Computer-aided quantitative bone scan assessment of prostate cancer treatment response.

Background: Cabozantinib is an oral inhibitor of MET and VEGFR2 and at a dose of 100 mg qd has demonstrated high rates of partial or complete bone scan resolution in CRPC patients with bone metastases, as assessed by manual visual reads. No accepted approach exists for the assessment of bone scan response in prostate cancer patients. An FDA 510(k) approved image analysis package has been developed that includes a computer-aided detection (CAD) system to detect pixels associated with bone lesions on bone scan with high accuracy, thus facilitating quantitative assessment of tumor burden with advantages of objectivity and consistency over manual approaches. A change in total Bone Scan Lesion Area of 30% or greater has been previously established as a cutoff point to distinguish responders from non-responders [1]. Methods: Original Digital Imaging and Communications in Medicine (DICOM) format images of baseline and week 6 bone scan assessments were collected and analyzed by the automated CAD system then reviewed by a physician experienced with bone scan interpretation as part of an ongoing clinical study of single agent cabozantinib at a lower starting dose of 40 mg qd in CRPC subjects with evidence of bone metastases. All subjects had undergone standard of care whole body scintigraphy, 2–4 hours post-injection of 20–25 mCi of Tc MDP. All images underwent intensity normalization to a bone scan reference to account for differences in radiotracer dosing levels and scan timing to improve reproducibility. Lesions consistent with metastatic foci of disease were automatically segmented based on anatomic region-specific intensity thresholds. Each segmented image was reviewed by a nuclear medicine physician for removal of any remaining false positive regions (e.g. areas of degenerative joint disease). Quantitative assessment of lesion burden was then determined. Results: 10 CRPC subjects showing evidence of bone metastasis were treated in an ongoing dose-ranging study with cabozantinib and were evaluable for bone scan response assessment at week 6. Reductions in the following parameters (median and range) were observed at week 6 relative to baseline: total Bone Scan Lesion Area (168.5 cm2 [20, 864] to 54 cm2 [0, 431]; 68% reduction), Bone Scan Lesion Count (34 [3, 74] to 3 [0, 64]; 91% reduction), and mean normalized Bone Scan Lesion Intensity (92.2 [56.5, 155.0] to 57.0 [0, 90.5]; 38% reduction). All 10 subjects had a reduction in Bone Scan Lesion Area, with 9 achieving a bone scan response at the week 6 timepoint as assessed by both the CAD system and visual review by the investigator. Conclusion: A significant reduction in apparent bone scan abnormality was demonstrated by the automated CAD system. The results demonstrate the potential for objective measurement of cabozantinib treatment effects in bone, laying the foundation for further validation against other clinically relevant outcome measures such as pain and overall survival. Reference: 1. G. H. Chu, M. S. Brown, H. J. Kim, M. Auerbach, C. Poon, B. Ramakrishna, A. Vidovic, D. W. Gjertson, M. J. Morris, S. M. Larson, J. G. Goldin, H. I. Scher. Initial analytic validation of automated bone scan measures for treatment response assessment in patients with metastatic castration-resistant prostate cancer (CRPC). J Clin Oncol 29: 2011 (suppl; abstr e15174). Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2011 Nov 12-16; San Francisco, CA. Philadelphia (PA): AACR; Mol Cancer Ther 2011;10(11 Suppl):Abstract nr B114.

Initial study of bone scan tumor area for early surrogate outcome assessment in patients with metastatic castration-resistant prostate cancer.

e15161 Background: Assessment of changes in tumor burden on bone scans is challenging. A computer-aided (CAD) system to measure overall tumor area on bone scans has been developed with pixel-level of 94.1% sensitivity 89.2% specificity, and 89.4% accuracy. We hypothesize that early changes in tumor area, determined by CAD, will be useful for predicting a patients progression rate. METHODS From an anonymized image database, 34 patients with castrate resistant prostate cancer undergoing one of two treatments (A and B) were randomly selected. Whole body standard of care scintigraphy was acquired in original DICOM format. Baseline and 14-week follow-up bone scans were analyzed with the CAD system. Early progressive disease (PD) was defined if the change in area was greater than 30% at 14 weeks. Multivariate Cox regression was use to test the PD effects on progression-free and patient survival rates, controlling for age and treatment. RESULTS At baseline, the median tumor area was 256 cm2 (267 cm2 for Treatment A and 250 cm2 for Treatment B) and 44% of patients were found to have PD at 14-week follow-up based on CAD assessment. Patients without PD had significantly longer progression-free beyond 14 weeks than subjects with PD: median 461 days vs. 135 days, p=0.04, hazard ratio= 3.0 (see table). The overall patient survival rates between non-PD and PD subjects were not statistically different (median 347 days vs. 284 days, p=0.42). CONCLUSIONS Early changes in computer-aided measurements of overall tumor area on bone scans holds promise as a surrogate outcome in prostate cancer studies and may aid in new drug development. [Table: see text].