Search

SignificanceOral cancer is one of the most prevalent cancers, especially in middle- and low-income countries such as India. Automatic segmentation of oral cancer images can improve the diagnostic workflow, which is a significant task in oral cancer image analysis. Despite the remarkable success of deep-learning networks in medical segmentation, they rarely provide uncertainty quantification for their output.AimWe aim to estimate uncertainty in a deep-learning approach to semantic segmentation of oral cancer images and to improve the accuracy and reliability of predictions.ApproachThis work introduced a UNet-based Bayesian deep-learning (BDL) model to segment potentially malignant and malignant lesion areas in the oral cavity. The model can quantify uncertainty in predictions. We also developed an efficient model that increased the inference speed, which is almost six times smaller and two times faster (inference speed) than the original UNet. The dataset in this study was collected using our customized screening platform and was annotated by oral oncology specialists.ResultsThe proposed approach achieved good segmentation performance as well as good uncertainty estimation performance. In the experiments, we observed an improvement in pixel accuracy and mean intersection over union by removing uncertain pixels. This result reflects that the model provided less accurate predictions in uncertain areas that may need more attention and further inspection. The experiments also showed that with some performance compromises, the efficient model reduced computation time and model size, which expands the potential for implementation on portable devices used in resource-limited settings.ConclusionsOur study demonstrates the UNet-based BDL model not only can perform potentially malignant and malignant oral lesion segmentation, but also can provide informative pixel-level uncertainty estimation. With this extra uncertainty information, the accuracy and reliability of the model's prediction can be improved.Oral cancer is one of the most prevalent cancers, especially in middle- and low-income countries such as India. Automatic segmentation of oral cancer images can improve the diagnostic workflow, which is a significant task in oral cancer image analysis. Despite the remarkable success of deep-learning networks in medical segmentation, they rarely provide uncertainty quantification for their output.We aim to estimate uncertainty in a deep-learning approach to semantic segmentation of oral cancer images and to improve the accuracy and reliability of predictions.This work introduced a UNet-based Bayesian deep-learning (BDL) model to segment potentially malignant and malignant lesion areas in the oral cavity. The model can quantify uncertainty in predictions. We also developed an efficient model that increased the inference speed, which is almost six times smaller and two times faster (inference speed) than the original UNet. The dataset in this study was collected using our customized screening platform and was annotated by oral oncology specialists.The proposed approach achieved good segmentation performance as well as good uncertainty estimation performance. In the experiments, we observed an improvement in pixel accuracy and mean intersection over union by removing uncertain pixels. This result reflects that the model provided less accurate predictions in uncertain areas that may need more attention and further inspection. The experiments also showed that with some performance compromises, the efficient model reduced computation time and model size, which expands the potential for implementation on portable devices used in resource-limited settings.Our study demonstrates the UNet-based BDL model not only can perform potentially malignant and malignant oral lesion segmentation, but also can provide informative pixel-level uncertainty estimation. With this extra uncertainty information, the accuracy and reliability of the model’s prediction can be improved.

To meet the increasing need for low-cost, compact imaging technology with cellular resolution, we have developed a microLED-based structured light sheet microscope for three-dimensional ex vivo and in vivo imaging of biological tissue in multiple modalities. All the illumination structure is generated directly at the microLED panel—which serves as the source—so light sheet scanning and modulation is completely digital, yielding a system that is simpler and less prone to error than previously reported methods. Volumetric images with optical sectioning are thus achieved in an inexpensive, compact form factor without any moving parts. We demonstrate the unique properties and general applicability of our technique by ex vivo imaging of porcine and murine tissue from the gastrointestinal tract, kidney, and brain.

Significance: Early detection of oral cancer is vital for high-risk patients, and machine learning-based automatic classification is ideal for disease screening. However, current datasets collected from high-risk populations are unbalanced and often have detrimental effects on the performance of classification. Aim: To reduce the class bias caused by data imbalance. Approach: We collected 3851 polarized white light cheek mucosa images using our customized oral cancer screening device. We use weight balancing, data augmentation, undersampling, focal loss, and ensemble methods to improve the neural network performance of oral cancer image classification with the imbalanced multi-class datasets captured from high-risk populations during oral cancer screening in low-resource settings. Results: By applying both data-level and algorithm-level approaches to the deep learning training process, the performance of the minority classes, which were difficult to distinguish at the beginning, has been improved. The accuracy of “premalignancy” class is also increased, which is ideal for screening applications. Conclusions: Experimental results show that the class bias induced by imbalanced oral cancer image datasets could be reduced using both data- and algorithm-level methods. Our study may provide an important basis for helping understand the influence of unbalanced datasets on oral cancer deep learning classifiers and how to mitigate.

In this Letter, a microLED-based chromatic confocal microscope with a virtual confocal slit is proposed and demonstrated for three-dimensional (3D) profiling without any mechanical scanning or external light source. In the proposed method, a micro-scale light-emitting diode (microLED) panel works as a point source array to achieve lateral scanning. Axial scanning is realized through the chromatic aberration of an aspherical objective. A virtual pinhole technique is utilized to improve the contrast and precision of depth reconstruction. The system performance has been demonstrated with a diamond-turned copper sample and onion epidermis. The experimental results show that the microLED panel could be a potential solution for portable 3D confocal microscopy. Several considerations and prospects are proposed for future microLED requirements in confocal imaging.

In medical imaging, deep learning-based solutions have achieved state-of-the-art performance. However, reliability restricts the integration of deep learning into practical medical workflows since conventional deep learning frameworks cannot quantitatively assess model uncertainty. In this work, we propose to address this shortcoming by utilizing a Bayesian deep network capable of estimating uncertainty to assess oral cancer image classification reliability. We evaluate the model using a large intraoral cheek mucosa image dataset captured using our customized device from high-risk population to show that meaningful uncertainty information can be produced. In addition, our experiments show improved accuracy by uncertainty-informed referral. The accuracy of retained data reaches roughly 90% when referring either 10% of all cases or referring cases whose uncertainty value is greater than 0.3. The performance can be further improved by referring more patients. The experiments show the model is capable of identifying difficult cases needing further inspection.

There are many excellent descriptions of mirativity in various language grammars, and more recently there has been a flurry of research refining mirativity to include how languages linguistically realize surprise and related concepts such as ‘unexpectedness’ and ‘new information’. However, there is currently no commonly accepted set of independently motivated diagnostics for testing mirativity that utilizes the best practices and first principles of semantic and pragmatic investigation. As such, the goal of this paper is to go back to basics and examine mirativity from the point of view of a field linguist who has been given the task of discovering and documenting how a speaker of a language linguistically expresses her surprise. This approach rests on two premises: first, mirativity is about surprise in the psychological sense. The second premise is that we take seriously that mirativity involves a kind of meaning, and that all languages have the linguistic resources for communicating mirative (surprise) meaning. The outcome is a set of tests that can be used to probe mirative meanings in any language.

In Urama, there are two clause-final particles, ka and ra , that encode a variety of both semantic and pragmatic meanings. While previous approaches have treated these particles as clause-type markers or evidential morphemes, this paper argues that one of these particles, ka , has another previously undocumented function in conversation: to mark speaker-knowledge and what the speaker assumes the addressee to know. We term these interactional uses of ka and ra . Functionally, the interactional use of ka follows from its clause-typing and speech act properties. Theoretically, Urama represents a language that has a grammatical strategy for tracking information in the Common Ground, which is close in spirit to evidentiality and clause-typing, but qualitatively different.

The aim of this paper is to investigate how evidentials are used to reflect the unprepared mind of the speaker. Lexical evidentials are typically used by speakers to encode the kinds of evidence they have for making statements about states, events and actions they did not personally witness or were a part of. However, under certain conditions evidentials can be used to express the surprise of the speaker. This is known in the literature under the label mirativity. The empirical factors that condition the mirative use of an evidential are first determined, and then using an information and schema-theoretic analysis it is shown that mirativity is the linguistic reflex of a series of mental events involving the processing of new information, coupled with other contextual factors involving speaker knowledge.

The goal of this paper is to investigate the linguistic, psychological and cognitive properties of utterances that express the surprise of the speaker, with a focus on how grammatical evidentials are used for this purpose. This is often labeled in the linguistics literature as mirativity. While there has been a flurry of recent interest in mirativity, we still lack an understanding of how and why evidentials are used this way, and an explanation of this effect. In this paper I take steps to filling this gap by showing how the mirativity associated with grammatical evidentials is one of the many linguistic reflexes of the more general cognitive process of surprise. I approach this by analyzing mirativity, and the language of surprise more generally, in a schema-theoretic framework enriched with the notion of new environmental information. I elaborate on the field methodological issues involved with testing the mirative use of an evidential and why they are used this way by connecting mirative evidentials to the broader phenomenon of deixis.

This chapter presents a sketch of the grammatical evidential system and related epistemic meanings in Gitksan, a critically endangered indigenous language of the Tsimshianic language family spoken in the northwest interior of Canada. A number of basic syntactic and semantic tests utilizing presupposition, negation, and dissent are applied that provide a nuanced description of the meanings of the individual evidentials. A specific feature of the Gitksan evidentials which is examined in detail involves how they can be used to express epistemic modality, and how a speaker’s choice of which evidential to use in a particular speech context is conditioned by her evaluation of the information acquired in that context. One of the effects of this choice is the expression of what can be translated as modal force.

This research examines the construction of masculinity among a group of working-class boys aged sixteen to nineteen in the northeast of England. Drawing on data collected from a six-week ethnography with boys in a religious (Christian) sixth form college, this study documents how only a small minority of these boys embodied the orthodox archetype of masculinity that has traditionally been associated with working-class youth. Instead, the great majority of participants adopted attitudes and behaviors that can be categorized as a set of inclusive masculinities: They espoused positive attitudes toward homosexuality, engaged in physical tactility and emotional intimacy, and used homosexually themed language without the intent to wound or marginalize other boys. These findings pose a considerable challenge to dominant narratives on working-class masculinities; narratives that must now be reconfigured to account for the proliferation of inclusive masculinities among working-class youth.

The Tsimshianic languages are entirely morphologically ergative in the agreement system. While there is a split in Tsimshianic, conditioned by both clause type and a person hierarchy, the other side of the split is not the expected nominative-accusative alignment. Rather, other logical groupings of semantic roles are found that are still ergative. This chapter presents a description of the agreement patterns across Tsimshianic, with the aim of explaining these expansions of ergativity, by undertaking a comparative analysis of the individual languages in the Tsimshianic family. This is analysis is extended to the connectives, which are complex, determiner-like morphemes that appear to be sensitive to the semantic role of the NP. This leads to four distinct alignments (nominative, ergative, neutral, and contrastive). An understanding of the alignments in the agreement system can shed light on this complexity, and a comparative analysis eliminates the multiple alignments in the connective system, thus revealing a fairly standard set of determiners.

In this article we examine the masculinities of heterosexual men in English university dance club settings. We highlight that multiple influences shape perceptions of gender and sexuality—influence...

Background— Although much is known about the risk factors for poor outcome in patients hospitalized with acute heart failure and left ventricular dysfunction, much less is known about the syndrome of acute heart failure primarily affecting the right ventricle (acute right heart failure). Methods and Results— By using Stanford Hospital's pulmonary hypertension database, we identified consecutive acute right heart failure hospitalizations in patients with PAH. We used longitudinal regression analysis with the generalized estimating equations method to identify factors associated with an increased likelihood of 90-day mortality or urgent transplantation. From June 1999 to September 2009, 119 patients with PAH were hospitalized for acute right heart failure (207 episodes). Death or urgent transplantation occurred in 34 patients by 90 days of admission. Multivariable analysis identified a higher respiratory rate on admission (>20 breaths per minute; OR, 3.4; 95% CI, 1.5–7.8), renal dysfunction on admission (glomerular filtration rate 2 ; OR, 2.7; 95% CI, 1.2–6.3), hyponatremia (serum sodium ≤136 mEq/L; OR, 3.6; 95% CI, 1.7–7.9), and tricuspid regurgitation severity (OR, 2.5 per grade; 95% CI, 1.2–5.5) as independent factors associated with an increased likelihood of death or urgent transplantation. Conclusions— These results highlight the high mortality after hospitalizations for acute right heart failure in patients with PAH. Factors identifiable within hours of hospitalization may help predict the likelihood of death or the need for urgent transplantation in patients with PAH.

This article revisits radical playgrounds of the past to offer a productive dialogue with recent debates on how child environments can foster citizenship and community. Joan Littlewood's playground projects are familiar examples of theatre techniques being applied to develop children's sense of belonging in a city. This essay considers the less familiar history of the Natural Theatre Company's Adventure Playground in Bath, an ambivalent site of chaotic transgression and community formation. Referencing the theories of Henri Lefebvre and Guy Debord, the essay explores how theatre techniques of subversive role playing, carnivalesque abandon and Do-It-Yourself (DIY) (de)construction can refigure notions of citizenship and community. With this article, I hope to trouble contemporary notions of safety by highlighting the restrictions placed on today's playgrounds and questioning whether forms of childhood development once valued can exist within the limitations of our modern ‘risk society’.

Background The assessment of pulmonary vascular resistance (PVR) plays an important role in the diagnosis and management of pulmonary arterial hypertension (PAH). The main objective of this study was to determine whether the noninvasive index of systolic pulmonary arterial pressure (SPAP) to heart rate (HR) times the right ventricular outflow tract time-velocity integral (TVI RVOT ) (SPAP/[HR × TVI RVOT ]) provides clinically useful estimations of PVR in PAH. Methods Doppler echocardiography and right-heart catheterization were performed in 51 consecutive patients with established PAH. The ratio of SPAP/(HR × TVI RVOT ) was then correlated with invasive indexed PVR (PVRI) using regression and Bland-Altman analysis. Using receiver operating characteristic curve analysis, a cutoff value for the Doppler equation was generated to identify patients with PVRI ≥ 15 Wood units (WU)/m 2 . Results The mean pulmonary arterial pressure was 52 ± 15 mm Hg, the mean cardiac index was 2.2 ± 0.6 L/min/m 2 , and the mean PVRI was 20.5 ± 9.6 WU/m 2 . The ratio of SPAP/(HR × TVI RVOT ) correlated very well with invasive PVRI measurements ( r = 0.860; 95% confidence interval, 0.759-0.920). A cutoff value of 0.076 provided well-balanced sensitivity (86%) and specificity (82%) to determine PVRI > 15 WU/m 2 . A cutoff value of 0.057 increased sensitivity to 97% and decreased specificity to 65%. Conclusion The novel index of SPAP/(HR × TVI RVOT ) provides useful estimations of PVRI in patients with PAH.

On the 17th floor of the Shard, I meditated on how the celebratory book launch of The Grammar of Politics and Performance might square with the environs of one the most emblematically corporatized ...

In this chapter, we take a different perspective on queer masculinities. Rather than examining the masculinities of gay men or queer-identifying individuals, we instead examine the masculinities of heterosexual men in a university setting. We highlight the multiple influences that shape perceptions of gender and sexuality—influences that are also used to subvert a polarized gender and sexuality order. This is evidenced by how straight men dance, interact, and even kiss each other. Accordingly, we ask what it means when queer masculinities are performed by otherwise straight-identifying men. What implications does the “homosexualization” of heterosexuals or the queering of straights have on understandings of gender and sexuality? We argue that, whether the context is a sporting event or a dance hall, social terrains rely on a body of assumed knowledge that helps construct the social meanings inculcated in and performed by moving bodies. Accordingly, in this chapter, we suggest that what used to be subversive signs of a polarized gender and sexuality order are increasingly found in the domain of popular and normative heterosexual culture. These social changes require a change in our theoretical formulations of masculinity.

Background Though much is known about the prognostic influence of acute kidney injury (AKI) in left-side heart failure, much less is known about AKI in patients with pulmonary arterial hypertension (PAH). Methods and Results We identified consecutive patients with PAH who were hospitalized at Stanford Hospital for acute right-side heart failure. AKI was diagnosed according to the criteria of the Acute Kidney Injury Network. From June 1999 to June 2009, 105 patients with PAH were hospitalized for acute right-side heart failure (184 hospitalizations). AKI occurred in 43 hospitalizations (23%) in 34 patients (32%). The odds of developing AKI were higher among patients with chronic kidney disease (odds ratio [OR] 3.9, 95% confidence interval [CI] 1.8–8.5), high central venous pressure (OR 1.8, 95% CI 1.1–2.4, per 5 mm Hg), and tachycardia on admission (OR 4.3, 95% CI 2.1–8.8). AKI was strongly associated with 30-day mortality after acute right-side heart failure hospitalization (OR 5.3, 95% CI 2.2–13.2). Conclusions AKI is relatively common in patients with PAH and associated with a short-term risk of death.

This paper reports the results of a research project that experiments with crosstabulation in aiding phonemic reconstruction. Data from the Tupi stock was used, and three tests were conducted in order to determine the efficacy of this application: the confirmation and challenging of a previously established reconstruction in the family; testing a new reconstruction generated by our model; and testing the upper limit of simultaneous, multiple correspondences across several languages. Our conclusion is that the use of cross tabulations (implemented within a database as pivot tables) offers an innovative and effective tool in comparative study and sound reconstruction.