An interactive learning framework for scalable classification of pathology images

Abstract

Recent advances in microscopy imaging and genomics have created an explosion of patient data in the pathology domain. Whole-slide images (WSIs) of tissues can now capture disease processes as they unfold in high resolution, recording the visual cues that have been the basis of pathologic diagnosis for over a century. Each WSI contains billions of pixels and up to a million or more microanatomic objects whose appearances hold important prognostic information. Computational image analysis enables the mining of massive WSI datasets to extract quantitative morphologic features describing the visual qualities of patient tissues. When combined with genomic and clinical variables, this quantitative information provides scientists and clinicians with insights into disease biology and patient outcomes. To facilitate interaction with this rich resource, we have developed a web-based machine-learning framework that enables users to rapidly build classifiers using an intuitive active learning process that minimizes data labeling effort. In this paper we describe the architecture and design of this system, and demonstrate its effectiveness through quantification of glioma brain tumors.

Visually-Enabled Active Deep Learning for (Geo) Text and Image Classification: A Review

“…Tuia et al [7] surveyed AL algorithms for RS image classification. Nalisink et al employed AL to reduce the labeling effort for image classification [8]. A good example using AL to overcome label quality problems by combining experts and crowd-sourced annotators can be found in [40].…”

“…The primary focus of research in M&DL has thus far been accurate results, often at the expense of human understanding of how the results were achieved [2][3][4][5][6]. However, accurate results often depend on building large human-generated training data sets that can be expensive in both financial and person cost to create [7][8][9][10][11][12][13]. As a result, there remain several impediments to broader adoption of M&DL, along with a range of concerns about potential negative outcomes related to the explainability of results produced.…”

Oracle or Teacher? A Systematic Overview of Research on Interactive Labeling for Machine Learning

“…Seven contributions in the cluster have the user select a subset to be labeled, use all data, or employ a hybrid approach. Hybrid approaches [25], combine user-and system-selected sampling. In particular, the user may provide self-chosen examples to augment the system-selected ones or be given the opportunity to correct both suggested labels by the system and introduce new labels or examples for existing labels.…”

“…In particular, the user may provide self-chosen examples to augment the system-selected ones or be given the opportunity to correct both suggested labels by the system and introduce new labels or examples for existing labels. For example, [25] work within the medical context to detect objects within microscopy images from pathological tissue samples. With their visualization approach, the authors enable the user to pick samples to label from a selection initially provided by the system.…”

Interactive phenotyping of large-scale histology imaging data with HistomicsML

“…We previously developed a basic software prototype to establish the feasibility of active learning classification with whole-slide imaging datasets 25 . This prototype developed important technology for visualizing whole-slide images and image analysis metadata via the web, but lacked critical features needed for dissemination as a tool and was not extensively validated.…”