Project description:While neural networks are used for classification tasks across domains, a long-standing open problem in machine learning is determining whether neural networks trained using standard procedures are consistent for classification, i.e., whether such models minimize the probability of misclassification for arbitrary data distributions. In this work, we identify and construct an explicit set of neural network classifiers that are consistent. Since effective neural networks in practice are typically both wide and deep, we analyze infinitely wide networks that are also infinitely deep. In particular, using the recent connection between infinitely wide neural networks and neural tangent kernels, we provide explicit activation functions that can be used to construct networks that achieve consistency. Interestingly, these activation functions are simple and easy to implement, yet differ from commonly used activations such as ReLU or sigmoid. More generally, we create a taxonomy of infinitely wide and deep networks and show that these models implement one of three well-known classifiers depending on the activation function used: 1) 1-nearest neighbor (model predictions are given by the label of the nearest training example); 2) majority vote (model predictions are given by the label of the class with the greatest representation in the training set); or 3) singular kernel classifiers (a set of classifiers containing those that achieve consistency). Our results highlight the benefit of using deep networks for classification tasks, in contrast to regression tasks, where excessive depth is harmful.
Project description:BackgroundThe incidence of melanoma is increasing faster than any other major cancer both in Brazil and worldwide. Southeast Brazil has especially high incidences of melanoma, and early detection is low. Exposure to ultraviolet (UV) radiation is a primary risk factor for developing melanoma. Increasing attractiveness is a major motivation among adolescents for tanning. A medical student-delivered intervention that takes advantage of the broad availability of mobile phones and adolescents' interest in their appearance indicated effectiveness in a recent study from Germany. However, the effect in a high-UV index country with a high melanoma prevalence and the capability of medical students to implement such an intervention remain unknown.ObjectiveIn this pilot study, our objective was to investigate the preliminary success and implementability of a photoaging intervention to prevent skin cancer in Brazilian adolescents.MethodsWe implemented a free photoaging mobile phone app (Sunface) in 15 secondary school classes in southeast Brazil. Medical students "mirrored" the pupils' altered 3-dimensional (3D) selfies reacting to touch on tablets via a projector in front of their whole grade accompanied by a brief discussion of means of UV protection. An anonymous questionnaire capturing sociodemographic data and risk factors for melanoma measured the perceptions of the intervention on 5-point Likert scales among 356 pupils of both sexes (13-19 years old; median age 16 years) in grades 8 to 12 of 2 secondary schools in Brazil.ResultsWe measured more than 90% agreement in both items that measured motivation to reduce UV exposure and only 5.6% disagreement: 322 (90.5%) agreed or strongly agreed that their 3D selfie motivated them to avoid using a tanning bed, and 321 (90.2%) that it motivated them to improve their sun protection; 20 pupils (5.6%) disagreed with both items. The perceived effect on motivation was higher in female pupils in both tanning bed avoidance (n=198, 92.6% agreement in females vs n=123, 87.2% agreement in males) and increased use of sun protection (n=197, 92.1% agreement in females vs n=123, 87.2% agreement in males) and independent of age or skin type. All medical students involved filled in a process evaluation revealing that they all perceived the intervention as effective and unproblematic, and that all pupils tried the app in their presence.ConclusionsThe photoaging intervention was effective in changing behavioral predictors for UV protection in Brazilian adolescents. The predictors measured indicated an even higher prospective effectiveness in southeast Brazil than in Germany (>90% agreement in Brazil vs >60% agreement in Germany to both items that measured motivation to reduce UV exposure) in accordance with the theory of planned behavior. Medical students are capable of complete implementation. A randomized controlled trial measuring prospective effects in Brazil is planned as a result of this study.
Project description:The exponential increase in algorithm-based mobile health (mHealth) applications (apps) for melanoma screening is a reaction to a growing market. However, the performance of available apps remains to be investigated. In this prospective study, we investigated the diagnostic accuracy of a class 1 CE-certified smartphone app in melanoma risk stratification and its patient and dermatologist satisfaction. Pigmented skin lesions ≥ 3 mm and any suspicious smaller lesions were assessed by the smartphone app SkinVision® (SkinVision® B.V., Amsterdam, the Netherlands, App-Version 6.8.1), 2D FotoFinder ATBM® master (FotoFinder ATBM® Systems GmbH, Bad Birnbach, Germany, Version 3.3.1.0), 3D Vectra® WB360 (Canfield Scientific, Parsippany, NJ, USA, Version 4.7.1) total body photography (TBP) devices, and dermatologists. The high-risk score of the smartphone app was compared with the two gold standards: histological diagnosis, or if not available, the combination of dermatologists', 2D and 3D risk assessments. A total of 1204 lesions among 114 patients (mean age 59 years; 51% females (55 patients at high-risk for developing a melanoma, 59 melanoma patients)) were included. The smartphone app's sensitivity, specificity, and area under the receiver operating characteristics (AUROC) varied between 41.3-83.3%, 60.0-82.9%, and 0.62-0.72% according to two study-defined reference standards. Additionally, all patients and dermatologists completed a newly created questionnaire for preference and trust of screening type. The smartphone app was rated as trustworthy by 36% (20/55) of patients at high-risk for melanoma, 49% (29/59) of melanoma patients, and 8.8% (10/114) of dermatologists. Most of the patients rated the 2D TBP imaging (93% (51/55) resp. 88% (52/59)) and the 3D TBP imaging (91% (50/55) resp. 90% (53/59)) as trustworthy. A skin cancer screening by combination of dermatologist and smartphone app was favored by only 1.8% (1/55) resp. 3.4% (2/59) of the patients; no patient preferred an assessment by a smartphone app alone. The diagnostic accuracy in clinical practice was not as reliable as previously advertised and the satisfaction with smartphone apps for melanoma risk stratification was scarce. MHealth apps might be a potential medium to increase awareness for melanoma screening in the lay population, but healthcare professionals and users should be alerted to the potential harm of over-detection and poor performance. In conclusion, we suggest further robust evidence-based evaluation before including market-approved apps in self-examination for public health benefits.
Project description:BackgroundThere is strong evidence suggesting that higher levels of cardiorespiratory fitness (CRF) are associated with a healthier metabolic profile, and that CRF can serve as a powerful predictor of morbidity and mortality. In this context, a smartphone app based on the 2-km walk test (UKK test) would provide the possibility to assess CRF remotely in individuals geographically distributed around a country or continent, and even between continents, with minimal equipment and low costs.ObjectiveThe overall aim of this study was to evaluate the validity and reliability of 2kmFIT-App developed for Android and iOS mobile operating systems to estimate maximum oxygen consumption (VO2max) as an indicator of CRF. The specific aims of the study were to determine the validity of 2kmFIT-App to track distance and calculate heart rate (HR).MethodsTwenty participants were included for field-testing validation and reliability analysis. The participants completed the UKK test twice using 2kmFIT-App. Distance and HR were measured with the app as well as with accurate methods, and VO2max was estimated using the UKK test equation.ResultsThe validity results showed the following mean differences (app minus criterion): distance (-70.40, SD 51.47 meters), time (-0.59, SD 0.45 minutes), HR (-16.75, SD 9.96 beats/minute), and VO2max (3.59, SD 2.01 ml/kg/min). There was moderate validity found for HR (intraclass correlation coefficient [ICC] 0.731, 95% CI -0.211 to 0.942) and good validity found for VO2max (ICC 0.878, 95% CI -0.125 to 0.972). The reliability results showed the following mean differences (retest minus test): app distance (25.99, SD 43.21 meters), app time (-0.15, SD 0.94 seconds), pace (-0.18, SD 0.33 min/km), app HR (-4.5, 13.44 beats/minute), and app VO2max (0.92, SD 3.04 ml/kg/min). There was good reliability for app HR (ICC 0.897, 95% CI 0.742-0.959) and excellent validity for app VO2max (ICC 0.932, 95% CI 0.830-0.973). All of these findings were observed when using the app with an Android operating system, whereas validity was poor when the app was used with iOS.ConclusionsThis study shows that 2kmFIT-App is a new, scientifically valid and reliable tool able to objectively and remotely estimate CRF, HR, and distance with an Android but not iOS mobile operating system. However, certain limitations such as the time required by 2kmFIT-App to calculate HR or the temperature environment should be considered when using the app.
Project description:This paper briefly reviews available technologies for measuring or estimating patient skin dose in the interventional fluoroscopic environment.
Project description:BackgroundA 9-grid scheme has been integrated into the foot and ankle literature to help clinicians and researchers localize osteochondral lesions of the talus (OLTs). We hypothesized that fellowship-trained orthopedic foot and ankle surgeons would have a high rate of intra/inter-observer reliability when localizing OLTs, therefore validating the scheme.MethodsWe queried our institution's foot and ankle radiographic database for magnetic resonance images with OLTs. Each MRI was reviewed by the senior author, and 2 key images (widest OLT diameter) from each tangential view were copied and combined onto one slide. Fifty consecutive deidentified images of ankles were then sent to 4 practicing fellowship-trained foot and ankle surgeons. Each was asked to identify which zone the OLT was localized within. A radiologist's report served as the control. Statistical analyses were performed using Cohen and Fleiss kappa tests.ResultsThe reviewers demonstrated majority consensus on 45/50 images with substantial agreement for zones 4 and 6. The interobserver reliability was moderate with a κ = 0.55. The mean intraobserver reliability was substantial, with a κ = 0.79. A musculoskeletal radiologist determined there were 3 lesions in zone 7, 18 lesions in zone 4, and 29 lesions in zone 6.ConclusionThis study is the first to critically evaluate the 9-grid scheme and its reliability among orthopedic foot and ankle surgeons. Our study found that the 9-grid scheme is an accurate method of localization for OLTs with high intra- and moderate interobserver reliability between surgeons.Level of evidenceLevel IV, retrospective diagnostic study.
Project description:Cross-cultural comparisons often investigate values that are assumed to have long-lasting influence on human conduct and thought. To capture and compare cultural values across cultures, Geert Hofstede's Cultural Dimensions Theory has offered an influential framework. Hofstede also provided a survey instrument, the Values Survey Module (VSM), for measuring cultural values as outlined in his Cultural Dimensions Theory. The VSM has since been subject to a series of revisions. Yet, data on countries have been derived from the original VSM - and not on one of the revised versions of VSM. We tested three scales (indulgence, power distance, and individualism) from the latest version, the VSM 2013, as part of a larger survey across 57 countries. Two main findings emerged. For one thing, country scores based on the VSM 2013 scales correlated only weakly with country scores of the same cultural dimensions obtained in a large previous study. Thus, the validity of the VSM 2013 is in doubt. For another thing, the internal consistency of the VSM 2013 scales was overall poor, indicating that the scales did not reliably measure well-defined constructs. We discuss implications for cross-cultural research.
Project description:Metamorphopsia, perceived as distortion of a shape, is experienced in age-related macular degeneration (AMD): straight lines appear to be curved and wavy to AMD patients and some other retinal pathologies. Conventional clinical assessment largely relies on asking patients to identify irregularities in Amsler Grids - a standardized set of equally spaced vertical and horizontal lines. Perceived distortions or gaps in the grid are a sign of macular pathology. Here, we developed an iterative Amsler Grid (IAG) procedure to obtain a quantifiable map of visual deformations. Horizontal and vertical line segments representing metamorphopsia are displayed on a computer screen. Line segments appearing distorted are adjusted by participants using the computer mouse to change their orientation in several iteratively such that they appear straight. Control participants are able to reliably correct deformations that simulate metamorphopsia while maintaining fixation in the center. In a pilot experiment, we attempted to obtain deformation maps from a small number of AMD patients. Whereas some patients with extensive scotomas found this procedure challenging, others were comfortable using the IAG and generating deformation maps corresponding to their subjective reports. This procedure may potentially be used to quantify local distortions and map them reliably in patients with early AMD.
Project description:The Haplopelma hainanum is a species of theraphosid spider from China. Its large size and charming appearance make this species a popular pet. According to a previous study, theraphosid spider bites can induce pain, erythema, and edema in humans and can present more severely in domestic animals. The pathological consequences of envenomation by H. hainanum remain unclear. In this study, we investigated the effects and mechanisms of H. hainanum envenomation in mice. We showed that the venom induced slight swelling, intense inflammatory response, and increased the microvascular density in mice skin. Moreover, we found that 50 µg/ml of the spider's venom induced IL-1? expression in both HaCaT cells and fibroblast cells, but repressed CXCL10 expression in fibroblasts. The venom significantly induced cell senescence and repressed cell proliferation and migration in both HaCaT cells and fibroblast cells. Finally, we examined the expression of Nav channel in HaCaT and fibroblast cells and found that H. hainanum venom effectively inhibited Na+ currents in HaCaT cells. Our study calls for further investigation of the pathological consequences and potential mechanisms of H. hainanum envenomation. This information might assist in the development of suitable therapy.