Remote smartphone monitoring of Parkinson's disease and individual response to therapy.

Remote health assessments that gather real-world data (RWD) outside clinic settings require a clear understanding of appropriate methods for data collection, quality assessment, analysis and interpretation. Here we examine the performance and limitations of smartphones in collecting RWD in the remote mPower observational study of Parkinson's disease (PD). Within the first 6 months of study commencement, 960 participants had enrolled and performed at least five self-administered active PD symptom assessments (speeded tapping, gait/balance, phonation or memory). Task performance, especially speeded tapping, was predictive of self-reported PD status (area under the receiver operating characteristic curve (AUC) = 0.8) and correlated with in-clinic evaluation of disease severity (r = 0.71; P < 1.8 × 10-6) when compared with motor Movement Disorder Society-Unified Parkinson's Disease Rating Scale (MDS-UPDRS). Although remote assessment requires careful consideration for accurate interpretation of RWD, our results support the use of smartphones and wearables in objective and personalized disease assessments.

Detecting the impact of subject characteristics on machine learning-based diagnostic applications.

Collection of high-dimensional, longitudinal digital health data has the potential to support a wide-variety of research and clinical applications including diagnostics and longitudinal health tracking. Algorithms that process these data and inform digital diagnostics are typically developed using training and test sets generated from multiple repeated measures collected across a set of individuals. However, the inclusion of repeated measurements is not always appropriately taken into account in the analytical evaluations of predictive performance. The assignment of repeated measurements from each individual to both the training and the test sets ("record-wise" data split) is a common practice and can lead to massive underestimation of the prediction error due to the presence of "identity confounding." In essence, these models learn to identify subjects, in addition to diagnostic signal. Here, we present a method that can be used to effectively calculate the amount of identity confounding learned by classifiers developed using a record-wise data split. By applying this method to several real datasets, we demonstrate that identity confounding is a serious issue in digital health studies and that record-wise data splits for machine learning- based applications need to be avoided.

Role of gp120 trimerization on HIV binding elucidated with Brownian adhesive dynamics.

We simulated the docking of human immunodeficiency virus (HIV) with a cell membrane using Brownian adhesive dynamics. The main advance in the current version of Brownian adhesive dynamics is that we use a simple bead-spring model to coarsely approximate the role of gp120 trimerization on HIV docking. We used our simulations to elucidate the effect of env spike density on the rate and probability of HIV binding, as well as the probability that each individual gp120 trimer is fully engaged. We found that for typical CD4 surface densities, viruses expressing as few as 8 env spikes will dock with binding rate constants comparable to viruses expressing 72 spikes. We investigated the role of cellular receptor diffusion on the degree of binding achieved by the virus on both short timescales (where binding has reached steady state but before substantial receptor accumulation in the viral-cell contact zone has occurred) and long timescales (where the system has reached steady state). On short timescales, viruses with 10-23 env trimers most efficiently form fully engaged trimers. On long timescales, all gp120 in the contact area will become bound to CD4. We found that it takes seconds for engaged trimers to cluster CD4 molecules in the contact zone, which partially explains the deleay in viral entry.

The Use of Smartphones for Health Research.

Because of their growing popularity and functionality, smartphones are increasingly valuable potential tools for health and medical research. Using ResearchKit, Apple's open-source platform to build applications ("apps") for smartphone research, collaborators have developed apps for researching asthma, breast cancer, cardiovascular disease, type 2 diabetes, and Parkinson disease. These research apps enhance widespread participation by removing geographical barriers to participation, provide novel ways to motivate healthy behaviors, facilitate high-frequency assessments, and enable more objective data collection. Although the studies have great potential, they also have notable limitations. These include selection bias, identity uncertainty, design limitations, retention, and privacy. As smartphone technology becomes increasingly available, researchers must recognize these factors to ensure that medical research is conducted appropriately. Despite these limitations, the future of smartphones in health research is bright. Their convenience grants unprecedented geographic freedom to researchers and participants alike and transforms the way clinical research can be conducted.

The Mole Mapper Study, mobile phone skin imaging and melanoma risk data collected using ResearchKit.

Sensor-embedded phones are an emerging facilitator for participant-driven research studies. Skin cancer research is particularly amenable to this approach, as phone cameras enable self-examination and documentation of mole abnormalities that may signal a progression towards melanoma. Aggregation and open sharing of this participant-collected data can be foundational for research and the development of early cancer detection tools. Here we describe data from Mole Mapper, an iPhone-based observational study built using the Apple ResearchKit framework. The Mole Mapper app was designed to collect participant-provided images and measurements of moles, together with demographic and behavioral information relating to melanoma risk. The study cohort includes 2,069 participants who contributed 1,920 demographic surveys, 3,274 mole measurements, and 2,422 curated mole images. Survey data recapitulates associations between melanoma and known demographic risks, with red hair as the most significant factor in this cohort. Participant-provided mole measurements indicate an average mole size of 3.95 mm. These data have been made available to engage researchers in a collaborative, multidisciplinary effort to better understand and prevent melanoma.

Mentorship Programs in Radiation Oncology Residency Training Programs: A Critical Unmet Need.

PURPOSE: To conduct a nationwide survey to evaluate the current status of resident mentorship in radiation oncology. METHODS AND MATERIALS: An anonymous electronic questionnaire was sent to all residents and recent graduates at US Accreditation Council for Graduate Medical Education-accredited radiation oncology residency programs, identified in the member directory of the Association of Residents in Radiation Oncology. Factors predictive of having a mentor and satisfaction with the mentorship experience were identified using univariate and multivariate analyses. RESULTS: The survey response rate was 25%, with 85% of respondents reporting that mentorship plays a critical role in residency training, whereas only 53% had a current mentor. Larger programs (≥ 10 faculty, P=.004; and ≥ 10 residents, P<.001) were more likely to offer a formal mentorship program, which makes it more likely for residents to have an active mentor (88% vs 44%). Residents in a formal mentoring program reported being more satisfied with the overall mentorship experience (univariate odds ratio 8.77, P<.001; multivariate odds ratio 5, P<.001). On multivariate analysis, women were less likely to be satisfied with the mentorship experience. CONCLUSIONS: This is the first survey focusing on the status of residency mentorship in radiation oncology. Our survey highlights the unmet need for mentorship in residency programs.

PERSONALIZED HYPOTHESIS TESTS FOR DETECTING MEDICATION RESPONSE IN PARKINSON DISEASE PATIENTS USING iPHONE SENSOR DATA.

We propose hypothesis tests for detecting dopaminergic medication response in Parkinson disease patients, using longitudinal sensor data collected by smartphones. The processed data is composed of multiple features extracted from active tapping tasks performed by the participant on a daily basis, before and after medication, over several months. Each extracted feature corresponds to a time series of measurements annotated according to whether the measurement was taken before or after the patient has taken his/her medication. Even though the data is longitudinal in nature, we show that simple hypothesis tests for detecting medication response, which ignore the serial correlation structure of the data, are still statistically valid, showing type I error rates at the nominal level. We propose two distinct personalized testing approaches. In the first, we combine multiple feature-specific tests into a single union-intersection test. In the second, we construct personalized classifiers of the before/after medication labels using all the extracted features of a given participant, and test the null hypothesis that the area under the receiver operating characteristic curve of the classifier is equal to 1/2. We compare the statistical power of the personalized classifier tests and personalized union-intersection tests in a simulation study, and illustrate the performance of the proposed tests using data from mPower Parkinsons disease study, recently launched as part of Apples ResearchKit mobile platform. Our results suggest that the personalized tests, which ignore the longitudinal aspect of the data, can perform well in real data analyses, suggesting they might be used as a sound baseline approach, to which more sophisticated methods can be compared to.

The mPower study, Parkinson disease mobile data collected using ResearchKit.

Current measures of health and disease are often insensitive, episodic, and subjective. Further, these measures generally are not designed to provide meaningful feedback to individuals. The impact of high-resolution activity data collected from mobile phones is only beginning to be explored. Here we present data from mPower, a clinical observational study about Parkinson disease conducted purely through an iPhone app interface. The study interrogated aspects of this movement disorder through surveys and frequent sensor-based recordings from participants with and without Parkinson disease. Benefitting from large enrollment and repeated measurements on many individuals, these data may help establish baseline variability of real-world activity measurement collected via mobile phones, and ultimately may lead to quantification of the ebbs-and-flows of Parkinson symptoms. App source code for these data collection modules are available through an open source license for use in studies of other conditions. We hope that releasing data contributed by engaged research participants will seed a new community of analysts working collaboratively on understanding mobile health data to advance human health.

Assessing the scale of tumor heterogeneity by complete hierarchical segmentation of MRI.

In many cancers, intratumoral heterogeneity has been found in histology, genetic variation and vascular structure. We developed an algorithm to interrogate different scales of heterogeneity using clinical imaging. We hypothesize that heterogeneity of perfusion at coarse scale may correlate with treatment resistance and propensity for disease recurrence. The algorithm recursively segments the tumor image into increasingly smaller regions. Each dividing line is chosen so as to maximize signal intensity difference between the two regions. This process continues until the tumor has been divided into single voxels, resulting in segments at multiple scales. For each scale, heterogeneity is measured by comparing each segmented region to the adjacent region and calculating the difference in signal intensity histograms. Using digital phantom images, we showed that the algorithm is robust to image artifacts and various tumor shapes. We then measured the primary tumor scales of contrast enhancement heterogeneity in MRI of 18 rhabdomyosarcoma patients. Using Cox proportional hazards regression, we explored the influence of heterogeneity parameters on relapse-free survival. Coarser scale of maximum signal intensity heterogeneity was prognostic of shorter survival (p = 0.05). By contrast, two fractal parameters and three Haralick texture features were not prognostic. In summary, our algorithm produces a biologically motivated segmentation of tumor regions and reports the amount of heterogeneity at various distance scales. If validated on a larger dataset, this prognostic imaging biomarker could be useful to identify patients at higher risk for recurrence and candidates for alternative treatment.

A patient-specific computational model of hypoxia-modulated radiation resistance in glioblastoma using 18F-FMISO-PET.

Glioblastoma multiforme (GBM) is a highly invasive primary brain tumour that has poor prognosis despite aggressive treatment. A hallmark of these tumours is diffuse invasion into the surrounding brain, necessitating a multi-modal treatment approach, including surgery, radiation and chemotherapy. We have previously demonstrated the ability of our model to predict radiographic response immediately following radiation therapy in individual GBM patients using a simplified geometry of the brain and theoretical radiation dose. Using only two pre-treatment magnetic resonance imaging scans, we calculate net rates of proliferation and invasion as well as radiation sensitivity for a patient's disease. Here, we present the application of our clinically targeted modelling approach to a single glioblastoma patient as a demonstration of our method. We apply our model in the full three-dimensional architecture of the brain to quantify the effects of regional resistance to radiation owing to hypoxia in vivo determined by [(18)F]-fluoromisonidazole positron emission tomography (FMISO-PET) and the patient-specific three-dimensional radiation treatment plan. Incorporation of hypoxia into our model with FMISO-PET increases the model-data agreement by an order of magnitude. This improvement was robust to our definition of hypoxia or the degree of radiation resistance quantified with the FMISO-PET image and our computational model, respectively. This work demonstrates a useful application of patient-specific modelling in personalized medicine and how mathematical modelling has the potential to unify multi-modality imaging and radiation treatment planning.

Periostin is a novel therapeutic target that predicts and regulates glioma malignancy.

BACKGROUND: Periostin is a secreted matricellular protein critical for epithelial-mesenchymal transition and carcinoma metastasis. In glioblastoma, it is highly upregulated compared with normal brain, and existing reports indicate potential prognostic and functional importance in glioma. However, the clinical implications of periostin expression and function related to its therapeutic potential have not been fully explored. METHODS: Periostin expression levels and patterns were examined in human glioma cells and tissues by quantitative real-time PCR and immunohistochemistry and correlated with glioma grade, type, recurrence, and survival. Functional assays determined the impact of altering periostin expression and function on cell invasion, migration, adhesion, and glioma stem cell activity and tumorigenicity. The prognostic and functional relevance of periostin and its associated genes were analyzed using the TCGA and REMBRANDT databases and paired recurrent glioma samples. RESULTS: Periostin expression levels correlated directly with tumor grade and recurrence, and inversely with survival, in all grades of adult human glioma. Stromal deposition of periostin was detected only in grade IV gliomas. Secreted periostin promoted glioma cell invasion and adhesion, and periostin knockdown markedly impaired survival of xenografted glioma stem cells. Interactions with αvß3 and αvß5 integrins promoted adhesion and migration, and periostin abrogated cytotoxicity of the αvß3/ß5 specific inhibitor cilengitide. Periostin-associated gene signatures, predominated by matrix and secreted proteins, corresponded to patient prognosis and functional motifs related to increased malignancy. CONCLUSION: Periostin is a robust marker of glioma malignancy and potential tumor recurrence. Abrogation of glioma stem cell tumorigenicity after periostin inhibition provides support for exploring the therapeutic impact of targeting periostin.

Prognostic value of primary tumor FDG uptake for occult mediastinal lymph node involvement in clinically N2/N3 node-negative non-small cell lung cancer.

OBJECTIVES: The objective of this study was to identify predictive factors of occult mediastinal nodal involvement on staging positron emission tomography with F-fluorodeoxyglucose in patients with non-small cell lung cancer. METHODS: We performed a retrospective review of 665 patients with suspected non-small cell lung cancer who underwent staging positron emission tomography with F-fluorodeoxyglucose from January 1, 2000 through August 31, 2010 at the Hospital of the University of Pennsylvania with clinical stage I or II disease and no evidence of N2 or N3 involvement on staging positron emission tomography (PET). A total of 201 of these patients underwent invasive pathologic staging of the mediastinum at the Hospital of the University of Pennsylvania with pathology reports available at the time of review. RESULTS: A total of 63 of the 201 patients were found to have N2 disease at the time of pathologic staging. The mean standardized uptake value (SUV) of the primary tumor for patients with occult N2 metastases was significantly higher than the node-negative patients (SUV 9.31 vs. 7.24, P=0.04). Histology, tumor location (central vs. peripheral), sex, and age were not predictive for occult N2 disease. A multivariate analysis was performed and identified primary tumor SUV>6 was the only significant predictor (P=0.02). An analysis by quartile identified a primary tumor SUV>10 to have an odds ratio of 1.72 compared with an SUV<4 of occult N2 involvement. CONCLUSIONS: Increased primary tumor SUV predicted for increased risk of mediastinal nodal disease. Tumor location was not predictive of PET-occult mediastinal nodal involvement, in contrast to previous publications. Pathologic staging of the mediastinum should be strongly considered in these patients even with a negative mediastinum on PET.

Invasion and proliferation kinetics in enhancing gliomas predict IDH1 mutation status.

BACKGROUND: Glioblastomas with a specific mutation in the isocitrate dehydrogenase 1 (IDH1) gene have a better prognosis than gliomas with wild-type IDH1. METHODS: Here we compare the IDH1 mutational status in 172 contrast-enhancing glioma patients with the invasion profile generated by a patient-specific mathematical model we developed based on MR imaging. RESULTS: We show that IDH1-mutated contrast-enhancing gliomas were relatively more invasive than wild-type IDH1 for all 172 contrast-enhancing gliomas as well as the subset of 158 histologically confirmed glioblastomas. The appearance of this relatively increased, model-predicted invasive profile appears to be determined more by a lower model-predicted net proliferation rate rather than an increased model-predicted dispersal rate of the glioma cells. Receiver operator curve analysis of the model-predicted MRI-based invasion profile revealed an area under the curve of 0.91, indicative of a predictive relationship. The robustness of this relationship was tested by cross-validation analysis of the invasion profile as a predictive metric for IDH1 status. CONCLUSIONS: The strong correlation between IDH1 mutation status and the MRI-based invasion profile suggests that use of our tumor growth model may lead to noninvasive clinical detection of IDH1 mutation status and thus lead to better treatment planning, particularly prior to surgical resection, for contrast-enhancing gliomas.

Patient-specific metrics of invasiveness reveal significant prognostic benefit of resection in a predictable subset of gliomas.

OBJECT: Malignant gliomas are incurable, primary brain neoplasms noted for their potential to extensively invade brain parenchyma. Current methods of clinical imaging do not elucidate the full extent of brain invasion, making it difficult to predict which, if any, patients are likely to benefit from gross total resection. Our goal was to apply a mathematical modeling approach to estimate the overall tumor invasiveness on a patient-by-patient basis and determine whether gross total resection would improve survival in patients with relatively less invasive gliomas. METHODS: In 243 patients presenting with contrast-enhancing gliomas, estimates of the relative invasiveness of each patient's tumor, in terms of the ratio of net proliferation rate of the glioma cells to their net dispersal rate, were derived by applying a patient-specific mathematical model to routine pretreatment MR imaging. The effect of varying degrees of extent of resection on overall survival was assessed for cohorts of patients grouped by tumor invasiveness. RESULTS: We demonstrate that patients with more diffuse tumors showed no survival benefit (P = 0.532) from gross total resection over subtotal/biopsy, while those with nodular (less diffuse) tumors showed a significant benefit (P = 0.00142) with a striking median survival benefit of over eight months compared to sub-totally resected tumors in the same cohort (an 80% improvement in survival time for GTR only seen for nodular tumors). CONCLUSIONS: These results suggest that our patient-specific, model-based estimates of tumor invasiveness have clinical utility in surgical decision making. Quantification of relative invasiveness assessed from routinely obtained pre-operative imaging provides a practical predictor of the benefit of gross total resection.

Impact of bioinformatic procedures in the development and translation of high-throughput molecular classifiers in oncology.

The progressive introduction of high-throughput molecular techniques in the clinic allows for the extensive and systematic exploration of multiple biologic layers of tumors. Molecular profiles and classifiers generated from these assays represent the foundation of what the National Academy describes as the future of "precision medicine". However, the analysis of such complex data requires the implementation of sophisticated bioinformatic and statistical procedures. It is critical that oncology practitioners be aware of the advantages and limitations of the methods used to generate classifiers to usher them into the clinic. This article uses publicly available expression data from patients with non-small cell lung cancer to first illustrate the challenges of experimental design and preprocessing of data before clinical application and highlights the challenges of high-dimensional statistical analysis. It provides a roadmap for the translation of such classifiers to clinical practice and makes key recommendations for good practice.

Toward patient-specific, biologically optimized radiation therapy plans for the treatment of glioblastoma.

PURPOSE: To demonstrate a method of generating patient-specific, biologically-guided radiotherapy dose plans and compare them to the standard-of-care protocol. METHODS AND MATERIALS: We integrated a patient-specific biomathematical model of glioma proliferation, invasion and radiotherapy with a multiobjective evolutionary algorithm for intensity-modulated radiation therapy optimization to construct individualized, biologically-guided plans for 11 glioblastoma patients. Patient-individualized, spherically-symmetric simulations of the standard-of-care and optimized plans were compared in terms of several biological metrics. RESULTS: The integrated model generated spatially non-uniform doses that, when compared to the standard-of-care protocol, resulted in a 67% to 93% decrease in equivalent uniform dose to normal tissue, while the therapeutic ratio, the ratio of tumor equivalent uniform dose to that of normal tissue, increased between 50% to 265%. Applying a novel metric of treatment response (Days Gained) to the patient-individualized simulation results predicted that the optimized plans would have a significant impact on delaying tumor progression, with increases from 21% to 105% for 9 of 11 patients. CONCLUSIONS: Patient-individualized simulations using the combination of a biomathematical model with an optimization algorithm for radiation therapy generated biologically-guided doses that decreased normal tissue EUD and increased therapeutic ratio with the potential to improve survival outcomes for treatment of glioblastoma.

Response classification based on a minimal model of glioblastoma growth is prognostic for clinical outcomes and distinguishes progression from pseudoprogression.

Glioblastoma multiforme is the most aggressive type of primary brain tumor. Glioblastoma growth dynamics vary widely across patients, making it difficult to accurately gauge their response to treatment. We developed a model-based metric of therapy response called Days Gained that accounts for this heterogeneity. Here, we show in 63 newly diagnosed patients with glioblastoma that Days Gained scores from a simple glioblastoma growth model computed at the time of the first postradiotherapy MRI scan are prognostic for time to tumor recurrence and overall patient survival. After radiation treatment, Days Gained also distinguished patients with pseudoprogression from those with true progression. Because Days Gained scores can be easily computed with routinely available clinical imaging devices, this model offers immediate potential to be used in ongoing prospective studies.

Discriminating survival outcomes in patients with glioblastoma using a simulation-based, patient-specific response metric.

Accurate clinical assessment of a patient's response to treatment for glioblastoma multiforme (GBM), the most malignant type of primary brain tumor, is undermined by the wide patient-to-patient variability in GBM dynamics and responsiveness to therapy. Using computational models that account for the unique geometry and kinetics of individual patients' tumors, we developed a method for assessing treatment response that discriminates progression-free and overall survival following therapy for GBM. Applying these models as untreated virtual controls, we generate a patient-specific "Days Gained" response metric that estimates the number of days a therapy delayed imageable tumor progression. We assessed treatment response in terms of Days Gained scores for 33 patients at the time of their first MRI scan following first-line radiation therapy. Based on Kaplan-Meier analyses, patients with Days Gained scores of 100 or more had improved progression-free survival, and patients with scores of 117 or more had improved overall survival. Our results demonstrate that the Days Gained response metric calculated at the routinely acquired first post-radiation treatment time point provides prognostic information regarding progression and survival outcomes. Applied prospectively, our model-based approach has the potential to improve GBM treatment by accounting for patient-to-patient heterogeneity in GBM dynamics and responses to therapy.