Skip to main content
eScholarship
Open Access Publications from the University of California

Scholarly Works (3 results)

  • Thesis
  • Peer Reviewed
UC San Francisco Electronic Theses and Dissertations (2019)

Two information technology revolutions are colliding in medicine. The first revolution has been the digitalization of health data, specifically Electronic Health Records (EHR). These records contain the details of who we are as patients, our ailments, treatments, and outcomes. Tragically, despite billions of dollars in investment from the US government, hardly any of this data is being utilized to better understand medicine or improve healthcare. This is largely because the data is voluminous, sparse, complex, and poorly formatted; making it unsuitable for traditional analytics methods. However the second revolution, modern Artificial Intelligence, specifically deep learning, provides tools, in the form of algorithms, to address exactly these problems. The primary difference between these modern algorithms and older ones is that the former are able to learn, more or less on their own, how to transform large complex data into a format that makes it easier to use and learn from.

In this dissertation, I have developed methods to apply deep learning to digital health data. Doing so, I have shown that we can predict the future health of individual patients with highly complex diseases, produced approaches to understand and leverage what these complex models are learning, and provided a framework for how healthcare systems of the near future could automatically learn to improve care daily.

For the first time in history, we are in a position to learn from the combined knowledge of tens of thousands of physicians and their experiences caring for hundreds of millions of patients. The potential transformations to healthcare are difficult to fully fathom, but certainly include safer, more powerful and efficient medicine, and a rapid speed up in new medical discoveries and treatments. Despite the promise, we must proceed carefully, balancing the great need to collectively use our data for better medicine with the individual right to privacy.

    Cover page: DEEP LEARNING IN PERSONALIZED MEDICINE: APPLICATIONS IN PATIENT SIMILARITY, PROGNOSIS, AND OPTIMAL TREATMENT SELECTION
    • Article
    • Peer Reviewed
    UC San Francisco Previously Published Works (2019)
    Applying state-of-the-art machine learning techniques to medical images requires a thorough selection and normalization of input data. One of such steps in digital mammography screening for breast cancer is the labeling and removal of special diagnostic views, in which diagnostic tools or magnification are applied to assist in assessment of suspicious initial findings. As a common task in medical informatics is prediction of disease and its stage, these special diagnostic views, which are only enriched among the cohort of diseased cases, will bias machine learning disease predictions. In order to automate this process, here, we develop a machine learning pipeline that utilizes both DICOM headers and images to predict such views in an automatic manner, allowing for their removal and the generation of unbiased datasets. We achieve AUC of 99.72% in predicting special mammogram views when combining both types of models. Finally, we apply these models to clean up a dataset of about 772,000 images with expected sensitivity of 99.0%. The pipeline presented in this paper can be applied to other datasets to obtain high-quality image sets suitable to train algorithms for disease detection.
      Cover page: Automatic Labeling of Special Diagnostic Mammography Views from Images and DICOM Headers
      • Article
      • Peer Reviewed
      UC San Francisco Previously Published Works (2020)
      There is a great and growing need to ascertain what exactly is the state of a patient, in terms of disease progression, actual care practices, pathology, adverse events, and much more, beyond the paucity of data available in structured medical record data. Ascertaining these harder-to-reach data elements is now critical for the accurate phenotyping of complex traits, detection of adverse outcomes, efficacy of off-label drug use, and longitudinal patient surveillance. Clinical notes often contain the most detailed and relevant digital information about individual patients, the nuances of their diseases, the treatment strategies selected by physicians, and the resulting outcomes. However, notes remain largely unused for research because they contain Protected Health Information (PHI), which is synonymous with individually identifying data. Previous clinical note de-identification approaches have been rigid and still too inaccurate to see any substantial real-world use, primarily because they have been trained with too small medical text corpora. To build a new de-identification tool, we created the largest manually annotated clinical note corpus for PHI and develop a customizable open-source de-identification software called Philter ("Protected Health Information filter"). Here we describe the design and evaluation of Philter, and show how it offers substantial real-world improvements over prior methods.
        Cover page: Protected Health Information filter (Philter): accurately and securely de-identifying free-text clinical notes
        Top