Physical Sciences and Mathematics Commons^™

Limitations Of Transformers On Clinical Text Classification, Shang Gao, Mohammed Alawad, Michael Todd Young, John Gounley, Noah Schaefferkoetter, Hong-Jun Yoon, Xiao-Cheng Wu, Eric B. Durbin, Jennifer Doherty, Antoinette Stroup, Linda Coyle, Georgia D. Tourassi

Kentucky Cancer Registry Faculty Publications

Bidirectional Encoder Representations from Transformers (BERT) and BERT-based approaches are the current state-of-the-art in many natural language processing (NLP) tasks; however, their application to document classification on long clinical texts is limited. In this work, we introduce four methods to scale BERT, which by default can only handle input sequences up to approximately 400 words long, to perform document classification on clinical texts several thousand words long. We compare these methods against two much simpler architectures -- a word-level convolutional neural network and a hierarchical self-attention network -- and show that BERT often cannot beat these simpler baselines when classifying …

Design, Synthesis, And Anticancer Properties Of Ru(Ii) Complexes With Organometallic, “Expanded” Bipyridine, And O,O’-Chelating Ligands, Raphael Ryan

Theses and Dissertations--Chemistry

Cancer is a worldwide public health crisis that requires new and improved drugs to be developed to extend survival rates and improve quality of life for the patient. Platinum-based drugs are used in approximately 50% of cancer treatment regimens. These drugs are highly effective in many kinds of cancer; however, cancers can develop platinum resistance and these drugs have troubling side effects that reduced their use and efficacy. To overcome these disadvantages, many other metals have been studied for their anticancer properties. Notably, the anticancer properties of ruthenium-based agents have drawn considerable attention with multiple ruthenium complexes entering clinical trials. …

Integrated Multiparametric Radiomics And Informatics System For Characterizing Breast Tumor Characteristics With The Oncotypedx Gene Assay, Michael A. Jacobs, Christopher B. Umbricht, Vishwa S. Parekh, Riham H. El Khouli, Leslie Cope, Katarzyna J. Macura, Susan Harvey, Antonio C. Wolff

Radiology Faculty Publications

Optimal use of multiparametric magnetic resonance imaging (mpMRI) can identify key MRI parameters and provide unique tissue signatures defining phenotypes of breast cancer. We have developed and implemented a new machine-learning informatic system, termed Informatics Radiomics Integration System (IRIS) that integrates clinical variables, derived from imaging and electronic medical health records (EHR) with multiparametric radiomics (mpRad) for identifying potential risk of local or systemic recurrence in breast cancer patients. We tested the model in patients (n = 80) who had Estrogen Receptor positive disease and underwent OncotypeDX gene testing, radiomic analysis, and breast mpMRI. The IRIS method was trained …

A Web-Based, Positive Emotion Skills Intervention For Enhancing Posttreatment Psychological Well-Being In Young Adult Cancer Survivors (Empower): Protocol For A Single-Arm Feasibility Trial, John M. Salsman, Laurie E. Mclouth, Michael Cohn, Janet A. Tooze, Mia Sorkin, Judith T. Moskowitz

Behavioral Science Faculty Publications

BACKGROUND: Adolescent and young adult cancer survivors (AYAs) experience clinically significant distress and have limited access to supportive care services. Interventions to enhance psychological well-being have improved positive affect and reduced depression in clinical and healthy populations but have not been routinely tested in AYAs.

OBJECTIVE: The aim of this protocol is to (1) test the feasibility and acceptability of a Web-based positive emotion skills intervention for posttreatment AYAs called Enhancing Management of Psychological Outcomes With Emotion Regulation (EMPOWER) and (2) examine proof of concept for reducing psychological distress and enhancing psychological well-being.

METHODS: The intervention development and testing are …

Using Case-Level Context To Classify Cancer Pathology Reports, Shang Gao, Mohammed Alawad, Noah Schaefferkoetter, Lynne Penberthy, Xiao-Cheng Wu, Eric B. Durbin, Linda Coyle, Arvind Ramanathan, Georgia Tourassi

Kentucky Cancer Registry Faculty Publications

Individual electronic health records (EHRs) and clinical reports are often part of a larger sequence-for example, a single patient may generate multiple reports over the trajectory of a disease. In applications such as cancer pathology reports, it is necessary not only to extract information from individual reports, but also to capture aggregate information regarding the entire cancer case based off case-level context from all reports in the sequence. In this paper, we introduce a simple modular add-on for capturing case-level context that is designed to be compatible with most existing deep learning architectures for text classification on individual reports. We …

Treatment And Outcomes Of Non-Small-Cell Lung Cancer Patients With High Comorbidity, Jorge Rios, Rahul Gosain, Bernardo H. L. Goulart, Bin Huang, Margaret N. Oechsli, Jaclyn K. Mcdowell, Quan Chen, Thomas Tucker, Goetz H. Kloecker

Biostatistics Faculty Publications

Background: The life expectancy of untreated non-small-cell lung cancer (NSCLC) is dismal, while treatment for NSCLC improves survival. The presence of comorbidities is thought to play a significant role in the decision to treat or not treat a given patient. We aim to evaluate the association of comorbidities with the survival of patients treated for NSCLC.

Methods: We performed a retrospective study of patients aged ≥66 years with invasive NSCLC between the years 2007 and 2011 in the Surveillance, Epidemiology, and End Results Kentucky Cancer Registry. Comorbidity was measured using the Klabunde Comorbidity Index (KCI), and univariate and multivariate logistic …

How Low Can You Go? Feature Selection For Drug Discovery, Derek Jones, Sally R. Ellingson, W. A. De Jong

Commonwealth Computational Summit

The cost of bringing a drug to market depends on how quickly a candidate drug can be “discovered” and evaluated to ensure safety and effectiveness. In this work we develop a method for predicting whether a given drug and protein compound will “bind.” Our aim is to select a set of features to predict drug-protein interactions.

This study focuses on kinases. Kinase inhibitors are the largest class of new cancer therapies. Selective inhibition is difficult due to high sequence similarity, leading to off-target interactions and side-effects. Pictured here human c-SRC.

Partner Interfering Behaviors Affecting Cancer Quality Of Life, Ann L. Coker, Diane R. Follingstad, Lisandra S. Garcia, Heather M. Bush

Obstetrics and Gynecology Faculty Publications

Objective: Because partners are an important and unpaid resource in cancer care, understanding how destructive, controlling or interfering partner behaviors influence women's cancer care may be particularly relevant for health care providers seeking to provide cancer care and enhance recovery. Using a new measure of partner interfering behaviors in cancer care (PIB-C), we investigated whether women with a recent cancer diagnosis who additionally endorsed any PIB-C would report (a) more symptoms of depression and stress, and (b) lower Functional Assessment of Cancer Therapy (FACT-G) and lower Functional Assessment of Chronic Illness Therapy-Spiritual Well-being (FACIT-SP) scale scores indicating poorer quality of …