Search

Weakly-supervised classification of histopathology slides is a computationally intensive task, with a typical whole slide image (WSI) containing billions of pixels to process. We propose Discriminative Region Active Sampling for Multiple Instance Learning (DRAS-MIL), a computationally efficient slide classification method using attention scores to focus sampling on highly discriminative regions. We apply this to the diagnosis of ovarian cancer histological subtypes, which is an essential part of the patient care pathway as different subtypes have different genetic and molecular profiles, treatment options, and patient outcomes. We use a dataset of 714 WSIs acquired from 147 epithelial ovarian cancer patients at Leeds Teaching Hospitals NHS Trust to distinguish the most common subtype, high-grade serous carcinoma, from the other four subtypes (low-grade serous, endometrioid, clear cell, and mucinous carcinomas) combined. We demonstrate that DRAS-MIL can achieve similar classification performance to exhaustive slide analysis, with a 3-fold cross-validated AUC of 0.8679 compared to 0.8781 with standard attention-based MIL classification. Our approach uses at most 18% as much memory as the standard approach, while taking 33% of the time when evaluating on a GPU and only 14% on a CPU alone. Reducing prediction time and memory requirements may benefit clinical deployment and the democratisation of AI, reducing the extent to which computational hardware limits end-user adoption.

Purpose - To characterise and assess the quality of published research evaluating artificial intelligence (AI) methods for ovarian cancer diagnosis or prognosis using histopathology data. Methods - A search of PubMed, Scopus, Web of Science, CENTRAL, and WHO-ICTRP was conducted up to 19/05/2023. The inclusion criteria required that research evaluated AI on histopathology images for diagnostic or prognostic inferences in ovarian cancer. The risk of bias was assessed using PROBAST. Information about each model of interest was tabulated and summary statistics were reported. PRISMA 2020 reporting guidelines were followed. Results - 1573 records were identified, of which 45 were eligible for inclusion. There were 80 models of interest, including 37 diagnostic models, 22 prognostic models, and 21 models with other diagnostically relevant outcomes. Models were developed using 1-1375 slides from 1-776 ovarian cancer patients. Model outcomes included treatment response (11/80), malignancy status (10/80), stain quantity (9/80), and histological subtype (7/80). All models were found to be at high or unclear risk of bias overall, with most research having a high risk of bias in the analysis and a lack of clarity regarding participants and predictors in the study. Research frequently suffered from insufficient reporting and limited validation using small sample sizes. Conclusion - Limited research has been conducted on the application of AI to histopathology images for diagnostic or prognostic purposes in ovarian cancer, and none of the associated models have been demonstrated to be ready for real-world implementation. Key aspects to help ensure clinical translation include more transparent and comprehensive reporting of data provenance and modelling approaches, as well as improved quantitative performance evaluation using cross-validation and external validations.

For many patients, current ovarian cancer treatments offer limited clinical benefit. For some therapies, it is not possible to predict patients' responses, potentially exposing them to the adverse effects of treatment without any therapeutic benefit. As part of the automated prediction of treatment effectiveness in ovarian cancer using histopathological images (ATEC23) challenge, we evaluated the effectiveness of deep learning to predict whether a course of treatment including the antiangiogenic drug bevacizumab could contribute to remission or prevent disease progression for at least 6 months in a set of 282 histopathology whole slide images (WSIs) from 78 ovarian cancer patients. Our approach used a pretrained Hierarchical Image Pyramid Transformer (HIPT) to extract region-level features and an attention-based multiple instance learning (ABMIL) model to aggregate features and classify whole slides. The optimal HIPT-ABMIL model had an internal balanced accuracy of 60.2% +- 2.9% and an AUC of 0.646 +- 0.033. Histopathology-specific model pretraining was found to be beneficial to classification performance, though hierarchical transformers were not, with a ResNet feature extractor achieving similar performance. Due to the dataset being small and highly heterogeneous, performance was variable across 5-fold cross-validation folds, and there were some extreme differences between validation and test set performance within folds. The model did not generalise well to tissue microarrays, with accuracy worse than random chance. It is not yet clear whether ovarian cancer WSIs contain information that can be used to accurately predict treatment response, with further validation using larger, higher-quality datasets required., Comment: Submission to ATEC23 challenge at MICCAI 2023 conference

Over a decade has gone by since the report of the Widening Participation Committee--"Learning works"--was presented to the Further Education Funding Council. Ten years on, this report asks how far have we come in the decade since "Learning works"? What progress have we made? And what barriers have we yet to overcome? Bringing together the perspectives of key thinkers, researchers and commentators, it aims to stimulate debate on what we need to do next to widen participation. (Contains 9 figures, 5 charts, 2 tables and 24 footnotes.) [This document was edited by Geoffrey Copland, Darshan Sachdev, and Colin Flint. The Foreword was written by Baroness Helena Kennedy.]

Since the emergence of COVID-19, deep learning models have been developed to identify COVID-19 from chest X-rays. With little to no direct access to hospital data, the AI community relies heavily on public data comprising numerous data sources. Model performance results have been exceptional when training and testing on open-source data, surpassing the reported capabilities of AI in pneumonia-detection prior to the COVID-19 outbreak. In this study impactful models are trained on a widely used open-source data and tested on an external test set and a hospital dataset, for the task of classifying chest X-rays into one of three classes: COVID-19, non-COVID pneumonia and no-pneumonia. Classification performance of the models investigated is evaluated through ROC curves, confusion matrices and standard classification metrics. Explainability modules are implemented to explore the image features most important to classification. Data analysis and model evaluations show that the popular open-source dataset COVIDx is not representative of the real clinical problem and that results from testing on this are inflated. Dependence on open-source data can leave models vulnerable to bias and confounding variables, requiring careful analysis to develop clinically useful/viable AI tools for COVID-19 detection in chest X-rays., Comment: To be published in MedInfo 21 - 18th World Congress on Medical and Health Informatics; 5 pages, 5 figures

In the manufacturing sector, the consideration of energy consumption during the scheduling and execution of jobs can offer significant benefits from an infrastructural and financial perspective. While numerous methods have been proposed for predicting the energy consumption of manufacturing machinery, they typically do not treat them as dynamic pieces of equipment which can lead to issues with long term accuracy. Furthermore, these models produce predictions at a high level of abstraction which can lead to sub-optimal utilization. This paper addresses these shortcomings and presents a new methodology based around the usage and inference of historical energy data. Multiple energy profiles for identical jobs are stored along with information regarding the machines mechanical conditions, allowing the system to compensate for machine-related changes to the energy consumption. Where historical data is lacking, analysis of how the machine's condition affects job energy consumption over time, allows for the use of Support Vector Regression to generate temporary synthetic energy profiles compensated for probable machine conditions.

In the manufacturing sector, energy demand is not considered as a manufacturing process variable when devising production schedules. This presents the potential for a large variance in the energy consumption culminating from the summation of individual machine energy demands. If not controlled, this can result in damage to the local power infrastructure. Traditional methods for protecting against this involve costly enhancements to the power infrastructure or inefficient use of time and equipment. In this paper, a production schedule modification algorithm is presented. Through the utilization of a genetic algorithm and highly granular historical energy profiles, the optimizer is able to modify an existing production schedule such that it produces a minimal variance in energy consumption when executed. Testing and experimentation show that a significant reduction in energy consumption variance can be achieved while ensuring the schedule operates within the constraints specified by the manufacturer.

In the manufacturing sector, energy demand is not considered as a manufacturing process variable when devising production schedules. This presents the potential for a large variance in the energy consumption culminating from the summation of individual machine energy demands. If not controlled, this can result in damage to the local power infrastructure. Traditional methods for protecting against this involve costly enhancements to the power infrastructure or inefficient use of time and equipment. In this paper, a production schedule modification algorithm is presented. Through the utilization of a genetic algorithm and highly granular historical energy profiles, the optimizer is able to modify an existing production schedule such that it produces a minimal variance in energy consumption when executed. Testing and experimentation show that a significant reduction in energy consumption variance can be achieved while ensuring the schedule operates within the constraints specified by the manufacturer.

The need for pavement preservation is widely recognized by transportation agencies. Over the past few years, the Maryland State Highway Administration (MDSHA) has made significant advancements in developing guidelines for pavement preservation in Maryland. The Pavement and Geotechnical Division (PAGD) within MDSHA has developed a pavement preservation guide that will assist in determining the "right fix for the right road at the right time" when used in conjunction with network-level and project-specific data. This guide was developed based on review of state-of-the-practice literature pertaining to pavement preservation, past experiences and the resource base of MDSHA pavement engineers. Step-by-step instructions on determining treatment options were developed through the use of flow charts, decision trees and treatment tables. At the end of the step-by-step process, there will be many viable treatment options available for consideration. In addition, various treatment options are also defined, along with information regarding treatment cost, advantages, disadvantages and other items. These options, along with the use of good judgment, common sense and research of current best practices help facilitate good decision-making. The ultimate selection of a preservation strategy by the District is based on the most cost-effective engineering design considering cost, practicality of construction, and benefit in terms of life extension. This paper presents an overview of the guide as of December 2014. Updates are made to this guide as new paving technologies emerge, and as feedback about performance of treatments is available. Efforts are also underway to integrate this guide into the pavement management optimization analysis.

Typical manufacturing scheduling algorithms do not consider the energy consumption of each job, or its variance, when they generate a production schedule. This can become problematic for manufacturers when local infrastructure has limited energy distribution capabilities. In this paper, a genetic algorithm based schedule modification algorithm is presented. By referencing energy consumption models for each job, adjustments are made to the original schedule so that it produces a minimal variance in the total energy consumption in a multi-process manufacturing production line, all while operating within the constraints of the manufacturing line and individual processes. Empirical results show a significant reduction in energy consumption variance can be achieved on schedules containing multiple concurrent jobs.

Appropriate social behaviour depends on the wiring together of relevant sensory and cognitive neural circuits during development and is impaired in Autism Spectrum Disorder (ASD). A functional variant of the MET gene at rs1858830 confers risk for ASD and a basic mechanistic hypothesis relating MET gene function to ASD has been proposed: Decreased MET protein expression during development increases the risk of ASD-relevant circuit miswiring. In the human, robust MET is localized to the posterior cingulate, inferior temporal, posterior parietal, and visual cortices, including face-processing regions. This pattern is consistent with the importance of vision in the social milieu of primates, suggesting a conserved developmental function of the MET receptor in facilitating the neural connectivity that allows for appropriate social behaviour. This study investigates the down-stream influence of the MET C allele at rs1858830 on morphological characteristics of brain regions involved in social processing of audiovisual information. We used an imaging-genetics approach to characterize neuroimaging intermediate phenotypes associated of the MET C allele in individuals with ASD (n= 14) and neurotypical controls. Audiovisual stimuli were presented in an emotion-matching task intended to engage diffuse networks of brain regions involved in cross-modal perceptual integration. We present modest evidence of an enhanced penetrance of the MET C risk allele in individuals with ASD, reflecting a mechanism whereby a common variant that is not disorder-causing in and of itself has a larger effect on network function in individuals with ASD compared to neurotypical individuals. Subtle differences between ASD and TD individuals were revealed when the cohort was stratified by MET risk genotype, demonstrating the potential power for an imaging-genetic approach to resolve phenotypic heterogeneity in ASD. Definition of neuroimaging interm, Master of Science (MSc)

Immersive visualisation is increasingly being used for comprehensive and rapid analysis of objects in 3D and object dynamic behaviour in 4D. Challenges are therefore presented to provide natural user interaction to enable effortless virtual object manipulation. Presented in this paper is the development and evaluation of an immersive human–computer interaction system based on stereoscopic viewing and natural hand gestures. For the development, it is based on the integration of a back-projection stereoscopic system for object and hand display, a hybrid inertial and ultrasonic tracking system to provide the absolute positions and orientations of the user’s head and hands, as well as a pair of high degrees-of-freedom data gloves to provide the relative positions and orientations of digit joints and tips on both hands. For the evaluation, it is based on a two-object scene with a virtual cube and a CT (computed tomography) volume created for demonstration of real-time immersive object manipulation. The system is shown to provide a correct user view of objects and hands in 3D with depth, as well as to enable a user to use a number of simple hand gestures to perform basic object manipulation tasks involving selection, release, translation, rotation and scaling. Also included in the evaluation are some quantitative tests of the system performance in terms of speed and latency.

Although SignWriting provides an intuitive notation system based on pictorial symbols to enable any sign based language in the world to be transcribed into a written form, it is a time consuming process for keyboard based input. To address the challenge of direct sign writing, the paper presents a human-computer-interaction system developed for recognition and visualisation of hand movements. The system is shown to be able to display the corresponding SignWriting symbols for various hand movements performed by two hands based on motion characteristics such as movement planes, movement directions, straight/curve movement paths, clockwise/anti-clockwise movements, and single/repeated movements.

Immersive visualisation is increasingly being used for comprehensive and rapid analysis of objects in 3D and object dynamic behaviour in 4D. Challenges are therefore presented to provide natural user interaction to enable effortless virtual object manipulation. Presented in this paper is the development and evaluation of a human-computer interaction system based on natural hand gestures. By employing a hybrid inertial and ultrasonic tracking system to provide the absolute positions and orientations of the userpsilas head and hands as well as a pair of high degrees-of-freedoms data gloves to provide the relative positions and orientations of finger joints and tips in both hands, the proposed system is shown to be able to automatically track and recognise a number of simple hand gestures. The effectiveness and potential of the proposed system is demonstrated through the five basic object manipulation tasks involving selection, release, translation, rotation and scaling of a 3D virtual cube.

The paper presents a new image based rendering method to enable a continuum of position and direction dependent views to be generated from a spherical panoramic image without apparent visual distortion, thereby allowing a user to immerse visually and wander freely within a spherical image environment. Presented in this paper include acquisition of spherical panorama from the real world, construction of a full spherical image environment by texture mapping of the spherical panorama onto a unit sphere, view generation based on viewing position and direction by using a virtual camera placed within the sphere, and correction of image distortion in the generated new views due to spherical surface curvature. Using an outdoor spherical panoramic image as an example, the quality of position and direction dependent views are shown, thereby demonstrating the potential of the proposed method for distortion free immersive navigation.

The Amigos two-way bilingual education program began as a collaborative effort between the Cambridge (Massachusetts) Public Schools' departments of desegregation and transitional bilingual education. Parents, teachers, administrators, and members of the community formed a committee to explore the possibility of developing a program that would combine the best features of transitional bilingual education (for limited-English-proficient students) and language immersion education (for native English speakers). The committee sought a way to end the isolation of language minority students from the rest of the school and to provide language majority students with the opportunity to acquire proficiency in a second language.The Amigos program commenced in September 1986 and currently serves close to 250 public school students, half of whom are from Spanish speaking homes, the other half of whom are from English-speaking homes. Half of their instruction is provided in Spanish, the other had in English.This report describes research that was conducted on the achievement in mathematics and in Spanish and English language arts of Amigos students and students in control/comparison groups. Also presented are data collected on students' and parents' attitudes toward bilingualism and biculturalism; students' self-assessment of academic competence and self-esteem; teachers' judgment of students' academic competence and self-esteem; and social-interactional patterns among Amigos students from different ethnic backgrounds.