Patterns最新文献

Most studies on temporal pattern visualization have focused on a single pattern and its metrics and supporting instances. However, the output of a mining process is typically an enumeration tree of frequent temporal patterns. A key challenge is exploring these patterns to identify those of interest for an expert or data scientist. Recently, it was suggested that the enumeration tree be browsed from the root downward through extended patterns. We introduce PanTeraV, a visualization system for statistical and analytical exploration of a large enumeration tree of complex temporal patterns. Demonstrated with time-interval-related patterns (TIRPs), it enables bidirectional exploration based on user-selected symbolic time intervals. The system consists of two visualizations: tabular, for navigating symbolic time intervals, and graphical, which presents relevant patterns in a bubble chart encoding multiple metrics. A user study on two real-world datasets shows that PanTeraV enables faster exploration of temporal patterns and allows users to discover associations of symbolic time intervals that were previously inaccessible.

The comprehensive histological assessment of chronic gastritis is imperative for guiding endoscopic follow-up strategies and surveillance of early-stage gastric cancer, yet rapid and objective assessment remains challenging in clinical workflows. We propose a powerful deep learning model (GastritisMIL) to effectively identify pathological alterations on H&E-stained biopsy slides, thereby expediting pathologists' evaluation and improving decision-making regarding follow-up intervals. We have trained and tested GastritisMIL by using retrospective data from 2,744 patients and evaluated discriminative performance across three medical centers (467 patients). GastritisMIL attained areas under the receiver operating curve greater than 0.971 in four tasks (inflammation, activity, atrophy, and intestinal metaplasia) and superior performance comparable to that of two senior pathologists. Specifically, interpretable attention heatmaps generated by GastritisMIL effectively assist junior pathologists in locating suspicious lesion regions across the entire field and minimizing missed diagnosis risk. Moreover, the high generalizability of this developed model across multiple external cohorts demonstrates its potential translational value.

DNA storage provides a high storage density and durability, but current systems experience high latency and lack data security. This study proposed the use of frequency dictionary mapping coding (FDMC) to enable handle-level random access of DNA archival storage. In addition, a hybrid e-molecular encryption strategy and a multi-level error-correction algorithm were introduced to ensure data security and integrity. The simulation and wet experiment results demonstrated that FDMC achieved handle-level random access in a lossless encrypted DNA storage system, which balanced security and robustness. Even in extreme cases, in which there was a 10% loss of DNA sequences, this system still recovered 91.74% of the original data while ensuring a storage density above 1.80 bits per nucleotide. In summary, FDMC enhances the application potential of DNA as a storage medium and bridges the gap between DNA storage and traditional storage modes.

Synthetic data are a popular method to publish useful datasets in a privacy-aware manner, making them useful across a range of scientific domains involving human subjects. They are typically generated by sampling from algorithms that mimic the probability distribution of real datasets, thereby maximizing statistical similarity to real data. However, we argue and demonstrate that synthetic data need to be similar only in ways relevant to their intended use and may neglect any irrelevant information, which in turn may improve privacy protection. As such, we propose a data synthesis method entitled fidelity-agnostic synthetic data. The method first extracts features relevant to the dataset's intended use using a neural net and then generates synthetic versions of the extracted features, after which they are decoded to mimic the real dataset. We show that our synthetic data improve performance in prediction tasks while retaining privacy protection compared to other state-of-the-art methods.

Artificial intelligence (AI) has the potential to greatly enhance diagnostic pathology, including the analysis of tissue samples to detect diseases such as colorectal cancer. This study explores how large language models (LLMs) and multimodal LLMs (MLLMs) can improve histopathological analysis by using medical data to aid diagnostics. However, challenges such as data quality and availability limit their effectiveness. To address these challenges, we introduce HistoChat, an AI-powered assistant designed to assist in colorectal cancer histopathology. It uses advanced techniques to improve data quality, such as generating image combinations and question-answer (QA) pairs to boost its learning. Despite working with limited data, HistoChat has significantly improved key metrics, including BLEU, ROUGE-L, and BERTScore, with an overall accuracy of 69.1% in human evaluation. These results suggest that HistoChat is a promising tool for enhancing diagnostic accuracy, especially in cases where data are scarce.

In liquid biopsy, detecting and differentiating circulating tumor cells (CTCs) and non-CTCs in metastatic cancer patients' blood samples remains challenging. The current gold standard often involves tedious manual examination of extensive image galleries. While machine learning (ML) offers potential automation, human expertise remains essential, particularly when ML systems face uncertainty or incorrect predictions due to limited labeled data. Combining self-supervised deep learning with an easily adaptable conventional ML classifier, we propose a human-in-the-loop approach with a targeted sampling strategy. By directing human efforts to label a limited set of new training samples from high-uncertainty clusters in the latent space, we iteratively reduce the system's uncertainty and improve classification performance, thereby saving time compared to naive sampling approaches. On data from metastatic breast cancer patients, we show the feasibility of our approach and achieve better performance while reducing expert evaluation time compared to the gold standard, the FDA-approved CellSearch system.

The QGIS project is a prominent open-source geographic information system (GIS) that has evolved over two decades, contributing significantly to the geospatial community. This paper presents the development, governance, and operational challenges faced by QGIS, providing an in-depth analysis of its growth from a hobby project to a global platform. We examine the project's organizational structure, release management, and infrastructure, alongside the financial model that sustains its development. The paper also addresses key challenges such as licensing complexities, group decision-making dynamics, and the balancing of innovation with stability in an open-source environment. Additionally, we highlight QGIS's broad applicability across industries and its continued success in fostering community-driven development.

Retrieving latent class patterns from complex data is challenging. This paper focuses on the problem of retrieving latent classes from local connections of raw data without any assumptions regarding data structures or distributions. We propose a framework called GULE (global understanding via local extraction) to address this challenge through both local extraction of class consistency and global propagation of the identified consistency. This paper provides a series of theoretical analyses to show why the GULE algorithm can retrieve latent classes with high accuracy. GULE can also serve as a tool for data visualization to preserve class topology structures. Comprehensive testing demonstrates that GULE provides precise clustering and highly reliable visualizations, potentially offering insights into diverse applications, including biology and medicine.

The exploration of efficient catalysts for sluggish sulfur redox reactions is pivotal for advancing lithium-sulfur batteries but remains inefficient through trial-and-error approaches. In a recent Joule study, Zhou, Li, and colleagues proposed an explainable-AI-based approach to intelligently design catalysts adaptive to diverse local chemical environments in batteries, achieving exceptional catalytic and battery performance.

In 2023, OpenAI's Democratic Inputs program funded 10 teams to design procedures for public participation in generative AI. In this perspective, we review the results of the project, drawing on interviews with some of the teams and our own experiences conducting participation exercises. We identify several shared yet largely unspoken assumptions of the project and encourage alternative forms of participation in AI perhaps coming from outside the tech industry.