To develop and validate a deep learning algorithm that predicts the final diagnosis of Alzheimer disease (AD), mild cognitive impairment, or neither at fluorine 18 (18 F) fluorodeoxyglucose (FDG) PET of the brain and compare its performance to that of radiologic readers. Materials and Methods: Prospective 18 F-FDG PET brain images from the Alzheimer's Disease Neuroimaging Initiative (ADNI) (2109 imaging studies from 2005 to 2017, 1002 patients) and retrospective independent test set (40 imaging studies from 2006 to 2016, 40 patients) were collected. Final clinical diagnosis at follow-up was recorded. Convolutional neural network of InceptionV3 architecture was trained on 90% of ADNI data set and tested on the remaining 10%, as well as the independent test set, with performance compared to radiologic readers. Model was analyzed with sensitivity, specificity, receiver operating characteristic (ROC), saliency map, and t-distributed stochastic neighbor embedding. Results: The algorithm achieved area under the ROC curve of 0.98 (95% confidence interval: 0.94, 1.00) when evaluated on predicting the final clinical diagnosis of AD in the independent test set (82% specificity at 100% sensitivity), an average of 75.8 months prior to the final diagnosis, which in ROC space outperformed reader performance (57% [four of seven] sensitivity, 91% [30 of 33] specificity; P , .05). Saliency map demonstrated attention to known areas of interest but with focus on the entire brain. Conclusion: By using fluorine 18 fluorodeoxyglucose PET of the brain, a deep learning algorithm developed for early prediction of Alzheimer disease achieved 82% specificity at 100% sensitivity, an average of 75.8 months prior to the final diagnosis.

The need to study dynamic biologic processes in intact smallanimal models of disease has stimulated the development of highresolution nuclear imaging methods. These methods are capable of clarifying molecular interactions important in the onset and progression of disease, assessing the biologic relevance of drug candidates and potential imaging agents, and monitoring therapeutic effectiveness of pharmaceuticals serially within a single-model system. Single-photon-emitting radionuclides have many advantages in these applications, and SPECT can provide 3-dimensional spatial distributions of g-(and x-) ray-emitting radionuclide imaging agents or therapeutics. Furthermore, combining SPECT with CT in a SPECT/CT system can assist in defining the anatomic context of biochemical processes and improve the quantitative accuracy of the SPECT data. Over the past decade, dedicated small-animal SPECT and SPECT/CT systems have been developed in academia and industry. Although significant progress in this arena has been realized through system development and biologic application, further innovation continues to address challenges in camera sensitivity, spatial resolution, and image reconstruction and quantification. The innumerable applications of small-animal SPECT and SPECT/CT in drug development, cardiology, neurology, and oncology are stimulating further investment in education, research, and development of these dedicated small-animal imaging modalities. Rapi dly evolving knowledge of molecular biology has stimulated exploration of novel therapies targeting specific points in molecular pathways associated with cardiac disease, neurologic disorders, cancer, and many other pathologic processes. However, identifying the role of a molecule in a disease process modeled in vitro does not necessarily translate to an understanding of its interactions with other molecular processes in vivo. On the other hand, few of all disease processes can be fully studied in human patients because of logistical and ethical concerns. Small-animal models represent a critical bridge between discoveries at the molecular level and implementation of clinically relevant diagnostics or therapeutics. Emphasis is ever increasing that these models accurately recapitulate both the disease itself and the environment in which the key molecular processes take place. For example, xenograft mouse models of cancer are simple to develop but are not considered particularly useful in understanding molecular interactions involved in carcinogenesis. More sophisticated approaches, such as transgenic models using oncogene activation or tumor suppressor inactivation, have evolved to the point at which cancers may be induced in a spatially and temporally defined manner using deletion of specified genetic sequences with Cre recombinase (1,2). Sophisticated infrastructures have been developed to manage data related to small-animal models of disease and provide greater access to various mouse models for all investigators, exemplified by the Mouse Models of Human Cancer Cons...

A major barrier to successful use of allogeneic hematopoietic cell transplantation is acute graft-versus-host disease (aGVHD), a devastating condition that arises when donor T cells attack host tissues. With current technologies, aGVHD diagnosis is typically made after end-organ injury and often requires invasive tests and tissue biopsies. This impacts patient prognosis as treatments are dramatically less effective at late disease stages. Here we show that a novel positron emission tomography (PET) radiotracer, 2′-deoxy-2′-[18F]fluoro-9-β-D-arabinofuranosylguanine ([18F]F-AraG), targeted towards two salvage kinase pathways preferentially accumulates in activated primary T cells. [18F]F-AraG PET imaging of a murine aGVHD model enabled visualization of secondary lymphoid organs harboring activated donor T cells prior to clinical symptoms. Tracer biodistribution in healthy humans showed favorable kinetics. This new PET strategy has great potential for early aGVHD diagnosis, enabling timely treatments and improved patient outcomes. [18F]F-AraG may be useful for imaging activated T cells in various biomedical applications.

131I-MIBG with myeloablative chemotherapy is feasible and effective for patients with neuroblastoma exhibiting de novo resistance to chemotherapy.

Semi-quantitative scoring of mIBG scans provides a more reliable method of assessing response in patients with relapsed neuroblastoma than qualitative impression. The reproducibility and high inter-observer concordance makes mIBG score an important component of overall response criteria in patients with recurrent neuroblastoma.

A B S T R A C T PurposeIodine-131-metaiodobenzylguanidine ( 131 I-MIBG) provides targeted radiotherapy with more than 30% response rate in refractory neuroblastoma, but activity infused is limited by radiation safety and hematologic toxicity. The goal was to determine the maximum-tolerated dose of 131 I-MIBG in two consecutive infusions at a 2-week interval, supported by autologous stem-cell rescue (ASCR) 2 weeks after the second dose. Patients and MethodsThe 131 I-MIBG dose was escalated using a 3 ϩ 3 phase I trial design, with levels calculated by cumulative red marrow radiation index (RMI) from both infusions. Using dosimetry, the second infusion was adjusted to achieve the target RMI, except at level 4, where the second infusion was capped at 21 mCi/kg. ResultsTwenty-one patients were enrolled onto the study at levels 1 to 4, with 18 patients assessable for toxicity and 20 patients assessable for response. Cumulative 131 I-MIBG given to achieve the target RMI ranged from 22 to 50 mCi/kg, with cumulative RMI of 3.2 to 8.92 Gy. No patient had a dose-limiting toxicity. Reversible grade 3 nonhematologic toxicity occurred in six patients at level 4, establishing the recommended cumulative dose as 36 mCi/kg. The median time to absolute neutrophil count more than 500/L after ASCR was 13 days (4 to 27 days) and to platelet independence was 17 days (6 to 47 days). Responses included two partial responses, eight mixed responses, three stable disease, and seven progressive disease. Responses by semiquantitative MIBG score occurred in eight patients, soft tissue responses occurred in five of 11 patients, but bone marrow responses occurred in only two of 13 patients. ConclusionThe lack of toxicity with this approach allowed dramatic dose intensification of 131 I-MIBG, with minimal toxicity and promising activity.

Delivery of therapeutics and imaging agents to target tissues requires localization and activation strategies with molecular specificity. Cell-associated proteases can be used for these purposes in a number of pathologic conditions, and their enzymatic activities can be exploited for activation strategies. Here, molecules based on the d-arginine octamer (r8) protein-transduction domain (PTD, also referred to as molecular transporters) have been adapted for selective uptake into cells only after proteolytic cleavage of a PTD-attenuating sequence by the prostate-specific antigen (PSA), an extracellular protease associated with the surface and microenvironment of certain prostate cancer cells. Convergent syntheses of these activatable PTDs (APTDs) are described, and the most effective r8 PTD-attenuating sequence is identified. The conjugates are shown to be stable in serum, cleaved by PSA, and taken up into Jurkat (human T cells) and PC3M prostate cancer cell lines only after cleavage by PSA. These APTD peptide-based molecules may facilitate targeted delivery of therapeutics or imaging agents to PSA-expressing prostate cancers.

The prognosis for patients diagnosed with mesothelioma is generally poor, and currently available treatments are usually ineffective. Therapies that specifically target tumor cells hold much promise for the treatment of cancers that are resistant to current approaches. We have previously selected phage antibody display libraries on mesothelioma cell lines to identify a panel of internalizing human single chain (scFv) antibodies that target mesothelioma-associated, clinically represented cell surface antigens and further exploited the internalizing function of these scFvs to specifically deliver lethal doses of liposome-encapsulated small molecule drugs to both epithelioid and sarcomatous subtypes of mesothelioma cells. Here, we report the identification of MCAM/MUC18/ CD146 as the surface antigen bound by one of the mesothelioma-targeting scFvs using a novel cloning strategy based on yeast surface human proteome display. Immunohistochemical analysis of mesothelioma tissue microarrays confirmed that MCAM is widely expressed by both epithelioid and sarcomatous types of mesothelioma tumor cells in situ but not by normal mesothelial cells. In addition, quantum dot-labeled anti-MCAM scFv targets primary meosthelioma cells in tumor fragment spheroids cultured ex vivo. As the first step in evaluating the therapeutic potential of MCAM-targeting antibodies, we performed single-photon emission computed tomography studies using the anti-MCAM scFv and found that it recognizes mesothelioma organotypic xenografts in vivo. The combination of phage antibody library selection on tumor cells and rapid target antigen identification by screening the yeast surface-displayed human proteome could be a powerful method for mapping the targetable tumor cell surface epitope space. [Cancer Res 2009;69(4):1570-7]