Diagnostic Review with Digital Pathology: Design of digitals tools for routine diagnostic use
Doctoral thesis, 2016
Digital pathology is a novel technology currently being implemented world
wide. This
thesis summarizes four years of HCI and visualization
research and provides an overall
understanding of designing
workstation software for pathologists. A human-centered
design
approach has been used to create a number of design interventions.
The thesis covers three main areas of inquiry: Understanding pathologists’
problem
solving processes during diagnostic review, how to build
different digital tools to support
those processes, and how to
incorporate digital image analysis algorithms when building
these
tools.
The thesis consist of a kappa that provides background and
context, to the remaining
appended papers. The papers describe
studies covering, pathologists’ navigation
strategies in gigapixel
sized images, the usability of different input devices and structured
reporting interfaces, how principles from volume rendering can be used for
multi-scale
images, and how make to use of machine learning
algorithms to support pathologists’
diagnostic processes.
Together, these design projects show how digital pathology images can be
used
to create tools to make pathologists more productive. This will
make it possible for
pathology laboratories to replace their
diagonstic workflow using glass slides, with a
workflow based
on digital images.
Digital pathology
Visualization
Human-computer interaction
Author
Jesper Molin
Chalmers, Applied Information Technology (Chalmers), Interaction design
Popular science description
English
Cancer tumors have been diagnosed for over 150 years by manually looking
at the tumor under the light of the microscope. New digital imaging technology
is now being developed to provide doctors with new powerful tools to enable
faster and more accurate cancer diagnostics. This way, today’s unnecessarily
long waiting times for cancer patients to receive their treatment could be
reduced.
Around 60 000 new cancer patients are diagnosed in Sweden every year. This means that every third person can expect to be diagnosed with cancer in their lifetime. However, treatments are getting better so two out of three cancer patients survive thanks to improved treatments. One of the most successful strategies has been to divide tumors into different subtypes and tailor the treatment for each specific tumor. Currently, cancer tumors are divided into around 200 subtypes. This strategy has put a tremendous burden on pathologists, the specialists that perform the subtyping.
This thesis presents the results from a research effort to create the next generation of digital diagnostic tools for pathologists. Instead of diagnosing the tumors with microscopes, they are scanned digitally, which allow diagnostic review using computer workstations. This makes it possible to use ergonomic input devices and to integrate report creation directly into the image review.
A further possibility that is explored is to use digital image analysis to automatically detect, classify and count tumor cells. This could help speed up the diagnostic process. However, tools based on automatic algorithms need to be used with care, since they cannot deal with unforeseen cases. The second part of the thesis explores ways that this problem could be overcome by using different interactive visualization techniques.
Around 60 000 new cancer patients are diagnosed in Sweden every year. This means that every third person can expect to be diagnosed with cancer in their lifetime. However, treatments are getting better so two out of three cancer patients survive thanks to improved treatments. One of the most successful strategies has been to divide tumors into different subtypes and tailor the treatment for each specific tumor. Currently, cancer tumors are divided into around 200 subtypes. This strategy has put a tremendous burden on pathologists, the specialists that perform the subtyping.
This thesis presents the results from a research effort to create the next generation of digital diagnostic tools for pathologists. Instead of diagnosing the tumors with microscopes, they are scanned digitally, which allow diagnostic review using computer workstations. This makes it possible to use ergonomic input devices and to integrate report creation directly into the image review.
A further possibility that is explored is to use digital image analysis to automatically detect, classify and count tumor cells. This could help speed up the diagnostic process. However, tools based on automatic algorithms need to be used with care, since they cannot deal with unforeseen cases. The second part of the thesis explores ways that this problem could be overcome by using different interactive visualization techniques.
Categorizing
Areas of Advance
Information and Communication Technology
Subject Categories (SSIF 2011)
Computer and Information Science
Design
Human Computer Interaction
Identifiers
ISBN
978-91-7597-497-2
Other
Series
Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie: 4178
Publisher
Chalmers
Public defence
2016-11-22 14:15
Jupiter 520
Opponent: Helwig Hauser, University of Bergen, Norway