PhD Dissertation Defense: Design and Development of a Clinical Decision Support System Using Machine Learning to Represent Domain Expertise in Wound Care Collaborative Decision Making

Design and Development of a Clinical Decision Support System Using Machine Learning to Represent Domain Expertise in Wound Care Collaborative Decision Making 

Abstract:  Many multi-disciplinary group decision-making scenarios are most effectively addressed with Collaborative Decision Making (CDM). Although CDM is critical for decision making quality, its feasibility diminishes when domain experts are unavailable, and decisions are time critical, as is the case in the chronic wound decisions, the focus of this dissertation. Decision Support systems (DSS) for clinical domain (also known as CDSS) that support CDM have been developed, but they require all collaborators to be available to contribute at the time of decision. 

The primary goal of this dissertation is to design, develop, evaluate, and test a Machine Learning (ML) model artifact that represents domain knowledge of expert collaborators in their absence and ensures accurate prediction of CDM, as a form of group decision making that involves consensus among multiple domain experts each of whose knowledge and experience is needed to make complex decisions. 

This dissertation utilizes the design science research (DSR) approach in each phase of ML workflow, i.e., data curation, feature selection, and model building. The resulting CDSS model for chronic wound care decisions takes (as inputs) wound features extracted from wound images and recommends (as the output) one of three standard of care pathways, i.e., to (1) continue current treatment, (2) request a change in treatment or (3) refer patient to a wound specialist clinic. In addition, this dissertation presents the usability evaluation of an instantiation of the ML-based CDSS artifact integrated into a smartphone app and tested with nurses (N= 4), who are its indented users. 

This dissertation contributes to the DSS literature by introducing a novel ML-based DSS artifact that facilitates CDM independent of expert availability and contributes to the CDSS literature by providing a solution for chronic wound care management.

Dissertation Committee:
Prof. Bengisu Tulu (Chair)
Prof. Diane M. Strong
Prof. Emmanuel O. Agu
Prof. Sharon Johnson

Location: Fuller Labs
Classroom 311

Zoom link
https://wpi.zoom.us/j/95704613201

Haadi Mombini
PhD Candidate

Haadi Mombini’s research interests are machine learning, deep learning, natural language processing, Explainable AI (XAI), and decision support systems (DSS). 

For his dissertation, Haadi developed and tested a machine learning model to be used as a reasoning engine for the “SmartWAnDS” wound decision-making app to help novice nurses manage chronic wound patients when access to wound specialty clinics or immediate specialist consultation is limited. This work has resulted in 3 conference proceedings and one journal publication.

Contact: hmombini@wpi.edu