Reproducibility in Radiomics: A Comparison of Feature Extraction Methods and Two Independent Datasets.

Radiomics involves the extraction of information from medical images that are not visible to the human eye. There is evidence that these features can be used for treatment stratification and outcome prediction. However, there is much discussion about the reproducibility of results between different studies. This paper studies the reproducibility of CT texture features used in radiomics, comparing two feature extraction implementations, namely the MATLAB toolkit and Pyradiomics, when applied to independent datasets of CT scans of patients: (i) the open access RIDER dataset containing a set of repeat CT scans taken 15 min apart for 31 patients (RIDER Scan 1 and Scan 2, respectively) treated for lung cancer; and (ii) the open access HN1 dataset containing 137 patients treated for head and neck cancer. Gross tumor volume (GTV), manually outlined by an experienced observer available on both datasets, was used. The 43 common radiomics features available in MATLAB and Pyradiomics were calculated using two intensity-level quantization methods with and without an intensity threshold. Cases were ranked for each feature for all combinations of quantization parameters, and the Spearman's rank coefficient, rs, calculated. Reproducibility was defined when a highly correlated feature in the RIDER dataset also correlated highly in the HN1 dataset, and vice versa. A total of 29 out of the 43 reported stable features were found to be highly reproducible between MATLAB and Pyradiomics implementations, having a consistently high correlation in rank ordering for RIDER Scan 1 and RIDER Scan 2 (rs > 0.8). 18/43 reported features were common in the RIDER and HN1 datasets, suggesting they may be agnostic to disease site. Useful radiomics features should be selected based on reproducibility. This study identified a set of features that meet this requirement and validated the methodology for evaluating reproducibility between datasets.

Patient-specific interactive software module for virtual preoperative planning and visualization of pedicle screw entry point and trajectories in spine surgery.

BACKGROUND: Lumbar pedicle screw insertion involves a steep learning curve for novice spine surgeons and requires image guidance or navigation. Small volume centers may be handicapped by the lack of cost-effective user-friendly tools for preoperative planning, guidance, and decision making. OBJECTIVE: We describe a patient-specific interactive software module, pedicle screw simulator (PSS), for virtual preoperative planning to determine the entry point and visualize the trajectories of pedicle screws. MATERIALS AND METHODS: The PSS was coded in Python for use in an open source image processing software, 3D Slicer. Preoperative computed tomography (CT) data of each subject was loaded into this module. The entry-target (ET) mode calculates the ideal angle from the entry point through the widest section of the pedicle to the desired target in the vertebral body. The entry-angle (EA) mode projects the screw trajectory from the desired entry point at a desired angle. The performance of this software was tested using CT data from four subjects. RESULTS: PSS provided a quantitative and qualitative feedback preoperatively to the surgeon about the entry point and trajectories of pedicle screws. It also enabled the surgeons to visualize and predict the pedicle breach with various trajectories. CONCLUSION: This interactive software module aids in understanding and correcting the orientation of each vertebra in three-dimensions, to identify the ideal entry points, angles of insertion and trajectories for pedicle screw insertion to suit the local anatomy.

Adaptive threshold segmentation of pituitary adenomas from FDG PET images for radiosurgery.

In this study we have attempted to optimize a PET based adaptive threshold seg- mentation method for delineating small tumors, particularly in a background of high tracer activity. The metabolic nature of pituitary adenomas and the constraints of MRI imaging in the postoperative setting to delineate these tumors during radio- surgical procedures motivated us to develop this method. Phantom experiments were done to establish a relationship between the threshold required for segmenting the PET images and the target size and the activity concentration within the target in relation to its background. The threshold was developed from multiple linear regression of the experimental data optimized for tumor sizes less than 4 cm3. We validated our method against the phantom target volumes with measured target to background ratios ranging from 1.6 to 14.58. The method was tested on ten retro- spective patients with residual growth hormone-secreting pituitary adenomas that underwent radiosurgery and compared against the volumes delineated by manual method. The predicted volumes against the true volume of the phantom inserts gave a correlation coefficient of 99% (p < 0.01). In the ten retrospective patients, the automatically segmented tumor volumes against volumes manually delineated by the clinicians had a correlation of 94% (p < 0.01). This adaptive threshold segmentation showed promising results in delineating tumor volumes in pituitary adenomas planned for stereotactic radiosurgery, particularly in the postoperative setting where MR and CT images may be associated with artifacts, provided opti- mization experiment is carried out. 

PET imaging of EGFR expression in nude mice bearing MDA-MB-468, a human breast adenocarcinoma.

BACKGROUND AND OBJECTIVE: Cetuximab is a monoclonal antibody that binds to and inhibits the epidermal growth factor receptor (EGFR). EGFR overexpression has been observed in a subset of breast cancers. The purpose of this study was to evaluate 64Cu-labeled cetuximab as an imaging agent using MDA-MB-468 breast cancer cells. METHODS: Cetuximab was coupled with an N-sulfosuccinimide ester of DOTA, purified, and labeled with the positron-emitting nuclide, 64Cu. Receptor-binding specificity and affinity of 64Cu-DOTA-cetuximab were studied using human MDA-MB-468 breast cancer cells, which express high levels of EGFR. Micropositron emission tomography and biodistribution studies were performed in athymic nude mice bearing MDA-MB-468 cell xenografts. Blocking studies with cold cetuximab were also performed to determine the specific binding of cetuximab. RESULTS: The radiochemical yield was 97.1 ± 1.1%. The specific activity was 1.5 Ci/µm cetuximab and the affinity to EGFR-positive MDA-MB-468 cells was high (KD=0.4 nmol/l). Both biodistribution and micropositron emission tomographic imaging studies with 64Cu-DOTA-cetuximab showed higher tumor uptake at 24 h (20.91 ± 2.49% ID/g, standardized uptake values of 9.6) than at 4 h (11.65 ± 3.89% ID/g, standardized uptake values of 4.9). Tumor uptake was significantly reduced from 20.91 ± 2.49% ID/g at 24 h to 14.42 ± 0.85% ID/g in a 1-h blocking study (P=0.00). CONCLUSION: Cetuximab can be labeled with 64Cu without compromising its biological activity. The tumor uptake was excellent with high tumor/muscle (7.97 ± 1.78 at 4 h, 15.91 ± 6.04 at 24 h) and reasonable tumor/blood (0.5 ± 0.18 at 4 h, 2.12 ± 0.86 at 24 h) ratios. Blocking studies showed the specific binding of the labeled antibody to tumor tissue.

Three dimensional projection environment for molecular design and surgical simulation.

We are developing agents for positron emission tomography (PET) imaging of cancer gene mRNA expression and software to fuse mRNA PET images with anatomical computerized tomography (CT) images to enable volumetric (3D) haptic (touch-and-feel) simulation of pancreatic cancer and surrounding organs prior to surgery in a particular patient. We have identified a novel ligand specific for epidermal growth factor receptor (EGFR) to direct PET agent uptake specifically into cancer cells, and created a volumetric haptic surgical simulation of human pancreatic cancer reconstructed from patient CT data. Young's modulus and the Poisson ratio for each tissue will be adjusted to fit the experience of participating surgeons.

Imaging spontaneous MMTVneu transgenic murine mammary tumors: targeting metabolic activity versus genetic products.

UNLABELLED: Despite the great strides made in imaging breast cancer (BC) in humans, the current imaging modalities miss up to 30% of BC, do not distinguish malignant lesions from benign ones, and require histologic examinations for which invasive biopsy must be performed. Annually in the United States, approximately 5.6 million biopsies find benign lesions. More than 50% of human BCs overexpress cyclin D1, and all BCs exhibit VPAC1 oncogene products. Together, these gene products may provide an excellent biomarker for the early and accurate detection of BC. We have evaluated 4 biologically active peptide analogs that have high affinity for VPAC1. The transgenic MMTVneu mice spontaneously develop BC and metastatic lesions that overexpress cyclin D1 and VPAC1 biomarkers. The MMTVneu mouse, therefore, provides an excellent animal model that mimics the pathogenesis of human BC. The objective of this investigation was to determine the ability of 1 of the peptide analogs, (64)Cu-TP3805, to detect BC in MMTVneu mice using (18)F-FDG as a gold standard. METHODS: The transgenic MMTVneu mouse colony was maintained. Offspring were screened for transgenic status by reverse transcriptase polymerase chain reaction (RT-PCR). Nine mice with visible, palpable, or unknown metastatic lesions were entered into the protocol. (18)F-FDG (6,475 +/- 1,628 kBq [175 +/- 44 microCi]) PET served as a control, followed by a CT scan and 24-48 h later by PET with (64)Cu-TP3805 (4,588 +/- 962 kBq [124 +/- 26 microCi]). RT-PCR on excised tumors determined VPAC1 expression, and histology ascertained the pathology. RESULTS: Ten tumors were detected by PET. Four tumors were detected both by (18)F-FDG and by (64)Cu-TP3805. Additionally, 4 tumors were imaged with (64)Cu-TP3805 only. These 8 tumors overexpressed VPAC1 receptors and were malignant by histology. The 2 remaining tumors were visualized with (18)F-FDG only. These tumors did not express the VPAC1 oncogene product and had benign histology. The standard uptake value ranged from 3.1 to 18.3 for (64)Cu-TP3805 and 0.9 to 1.4 for (18)F-FDG. CONCLUSION: (64)Cu-TP3805 identified all malignant lesions unequivocally that overexpressed the VPAC1 oncogene surface product. The 2 benign tumors that did not express the VPAC1 receptor were not imaged. (64)Cu-TP3805 promises to have the potential for the early and accurate imaging of primary and metastatic BC.

In-house preparation of iodine -131 metaiodo benzyl guanidine for scintigraphy of neuroendocrine tumors. Fourteen years experience in South India.

Iodine-131 metaiodobenzyl guanidine ((131)I-MIBG) is routinely used for imaging and treatment of neuroendocrine tumors (NET). As the commercially available radiopharmaceutical was very expensive, we developed an in-house method of labeling MIBG with (131)I in 1993. A total of 247 batches of (131)I-MIBG were prepared and used in our hospital between April 1993 and September 2006. We report our experience over these 14 years of preparation of this tracer in our hospital radiopharmacy, for the scintigraphy of NET. The technique of preparation is simple and the labeled product was found to be of acceptable quality. With the routine availability and cost effectiveness, the utilization of this radiopharmaceutical for scintigraphy increased remarkably in our institution.