Grant: Master's Award: Canada Graduate Research Scholarships - Master's program
University of Toronto
$27,000.00
Medical Image Synthesis for Expediting the Diagnosis to Treatment Pathway in Radiation Oncology
FY 2025-26
Summary
A major source of the delay between cancer diagnosis and the start of radiation therapy (RT) is the need for a separate imaging session specifically for planning RT treatments. This typically delays treatment initiation by many days to week following diagnosis, which results in poorer patient outcomes. This project aims to develop a machine learning (ML) model that synthesizes RT planning images from already available diagnostic images, reducing the time from diagnosis to treatment and streamlining the RT workflow. The benefits of this project to patient care include reducing wait times, streamlining clinical resources use, decreasing the number of hospital visits, and enhancing treatment quality and outcomes through earlier initiation. The ML model will be trained using paired RT planning and diagnostic CTs acquired for about 130 prostate and 230 head and neck cancer patients. Generative ML architectures will be used to synthesize RT planning CTs from diagnostic CTs, correcting for differences in patient positioning, imaging protocols, and acquisition parameters, enabling diagnostic images to be used for RT treatment planning. Model performance will be evaluated using standard image quality metrics, including peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM). Additionally, task-based evaluation will be completed by segmenting critical anatomical structures to assess geometric accuracy, and dosimetric evaluation will be performed by generating RT treatment plans on the synthetic planning CT images and comparing them to those created from conventional planning CTs.
Department
Program
Master's Award: Canada Graduate Research Scholarships - Master's program
Recipient
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
CIHR Funding Decisions Database
As published
Id
572589
Name
Lollino Sophia
Region
Ontario
Country
Canada
Orgtype
Institution Paid; Research Institution
Prcname2
Special Cases - Awards Programs
Projecttitle
Medical Image Synthesis for Expediting the Diagnosis to Treatment Pathway in Radiation Oncology
Programtype2
Studentship Programs
Orgnameinp2
University of Toronto
Cihramount2
$27,000
Cihrequipment2
$0
Programname2
Master's Award: Canada Graduate Research Scholarships - Master's program
Keyworddelim
Artificial Intelligence; Diagnostic Computed Tomography; Image Synthesis; Image Translation; Machine Learning; Medical Physics; Radiation Oncology; Radiation Therapy; Simulation Computed Tomography; Treatment Planning
Abstract
A major source of the delay between cancer diagnosis and the start of radiation therapy (RT) is the need for a separate imaging session specifically for planning RT treatments. This typically delays treatment initiation by many days to week following diagnosis, which results in poorer patient outcomes. This project aims to develop a machine learning (ML) model that synthesizes RT planning images from already available diagnostic images, reducing the time from diagnosis to treatment and streamlining the RT workflow. The benefits of this project to patient care include reducing wait times, streamlining clinical resources use, decreasing the number of hospital visits, and enhancing treatment quality and outcomes through earlier initiation. The ML model will be trained using paired RT planning and diagnostic CTs acquired for about 130 prostate and 230 head and neck cancer patients. Generative ML architectures will be used to synthesize RT planning CTs from diagnostic CTs, correcting for differences in patient positioning, imaging protocols, and acquisition parameters, enabling diagnostic images to be used for RT treatment planning. Model performance will be evaluated using standard image quality metrics, including peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM). Additionally, task-based evaluation will be completed by segmenting critical anatomical structures to assess geometric accuracy, and dosimetric evaluation will be performed by generating RT treatment plans on the synthetic planning CT images and comparing them to those created from conventional planning CTs.
Competitiondate
202512
Approvedterm2
1 yr 0 mth
Pinamesdelim
Lollino, Sophia
Approvedterm
1 an 0 mois
Cihrcontribution2
~ $27,000~ $0~