Adaptive Dimensionality Reduction with Semi-Supervision (AdDReSS): Classifying Multi-Attribute Biomedical Data

• Published in: PloS one Vol. 11; no. 7; p. e0159088
• Main Authors: Lee, George, Romo Bucheli, David Edmundo, Madabhushi, Anant
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 15.07.2016; Public Library of Science (PLoS)
• Subjects: Active learning; Algorithms; Artificial Intelligence; Biology and Life Sciences; Biomedical data; Biomedical engineering; Brain; Breast - pathology; Breast Neoplasms - diagnosis; Classification; Computer and Information Sciences; Data mining; Data processing; Diagnosis, Computer-Assisted - methods; Diagnostic imaging; Dimensional analysis; Disease; Embedding; Female; Gene expression; Gene Expression Regulation, Neoplastic; Histopathology; Humans; Image Processing, Computer-Assisted - methods; Learning; Magnetic resonance; Magnetic resonance imaging; Male; Medicine and Health Sciences; Methods; Mitosis; Neuroimaging; NMR; Nuclear magnetic resonance; Ovarian Neoplasms - diagnosis; Ovary - pathology; Pattern Recognition, Automated - methods; Physical Sciences; Principal components analysis; Prostate; Prostate - pathology; Prostate cancer; Prostatic Neoplasms - diagnosis; Prostatic Neoplasms - genetics; Proteomics; Random sampling; Reduction; Representations; Research and Analysis Methods; Statistical sampling; Colombia; Cleveland Ohio; United States > US
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0159088

Medical diagnostics is often a multi-attribute problem, necessitating sophisticated tools for analyzing high-dimensional biomedical data. Mining this data often results in two crucial bottlenecks: 1) high dimensionality of features used to represent rich biological data and 2) small amounts of labelled training data due to the expense of consulting highly specific medical expertise necessary to assess each study. Currently, no approach that we are aware of has attempted to use active learning in the context of dimensionality reduction approaches for improving the construction of low dimensional representations. We present our novel methodology, AdDReSS (Adaptive Dimensionality Reduction with Semi-Supervision), to demonstrate that fewer labeled instances identified via AL in embedding space are needed for creating a more discriminative embedding representation compared to randomly selected instances. We tested our methodology on a wide variety of domains ranging from prostate gene expression, ovarian proteomic spectra, brain magnetic resonance imaging, and breast histopathology. Across these various high dimensional biomedical datasets with 100+ observations each and all parameters considered, the median classification accuracy across all experiments showed AdDReSS (88.7%) to outperform SSAGE, a SSDR method using random sampling (85.5%), and Graph Embedding (81.5%). Furthermore, we found that embeddings generated via AdDReSS achieved a mean 35.95% improvement in Raghavan efficiency, a measure of learning rate, over SSAGE. Our results demonstrate the value of AdDReSS to provide low dimensional representations of high dimensional biomedical data while achieving higher classification rates with fewer labelled examples as compared to without active learning.