Development of Machine Learning Methodologies to Improve Outcomes in Chronic Inflammatory Diseases
Başlık:
Development of Machine Learning Methodologies to Improve Outcomes in Chronic Inflammatory Diseases
Yazar:
Konkayala, Bhargava Reddy, author.
ISBN:
9780438066236
Yazar Ek Girişi:
Fiziksel Tanımlama:
1 electronic resource (89 pages)
Genel Not:
Source: Dissertation Abstracts International, Volume: 79-10(E), Section: A.
Advisors: Dursun Delen Committee members: Bruce Benjamin; Toby Joplin; Ramesh Sharda.
Özet:
The health care industry is going through significant changes in the delivery of care and reimbursements. These changes coupled with advancements in technology will improve patient outcomes by predicting an event before it actually occurs. Although recent literature is quite rich in data analytics solutions for chronic diseases (congestive heart failure, diabetes, cancer, hypertension, etc.), very little has been done to study rare chronic inflammatory diseases. This may be due to of the complexity of the diseases and a lack of thorough understanding of the disease pathways and fewer clinical trials conducted to study rare diseases in depth. This research has two essays from the health care domain covering the rare disease areas using EMR (electronic medical records) data. Data from patients diagnosed with Crohn's disease and lupus are utilized for the essays.
In the first essay (Chapter III), I design, develop, and validated three machine-learning methodologies---logistic regression, regularized or penalized logistic regression, and gradient boosting machines---to predict inflammation in Crohn's disease patients using EMR data. Results show that machine learning algorithms, gradient boosting machines, can predict inflammation severity with 92.82% accuracy, followed by regularized regression with 82.70%, and logistic regression with 81.12%. A combination of baseline laboratory parameters, patient demographic characteristics, and disease location are found to be the strongest predictors of the severity of inflammation.
In the second essay (Chapter IV), I conduct research to design, evaluate, and identify best performing models, from both traditional classification models and novel deep learning methodologies to predict hospital readmission. Performance of traditional machine learning methods---penalized logistic regression, artificial neural networks, and deep learning methods such as long short term memory (LSTM)---are evaluated and compared. Results show that LSTM has significantly better performance with an AUC of 0.70 compared to traditional classification methods such as ANN and penalized logistic regression with AUCs of 0.66 and 0.63, respectively. This superior performance by the deep learning method may be due to the ability of the deep learning models to capture and leverage temporal relationships between visits and progression of the disease over time.
Notlar:
School code: 0664
Tüzel Kişi Ek Girişi:
Mevcut:*
Yer Numarası | Demirbaş Numarası | Shelf Location | Lokasyon / Statüsü / İade Tarihi |
---|---|---|---|
XX(678047.1) | 678047-1001 | Proquest E-Tez Koleksiyonu | Arıyor... |
On Order
Liste seç
Bunu varsayılan liste yap.
Öğeler başarıyla eklendi
Öğeler eklenirken hata oldu. Lütfen tekrar deneyiniz.
:
Select An Item
Data usage warning: You will receive one text message for each title you selected.
Standard text messaging rates apply.