Download Poster - CRCV - University of Central Florida

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Preventive healthcare wikipedia , lookup

Transcript
Diagnosing Lung Cancer with Computer-Aided Detection:
Role-Playing in the Advanced Placement Biology Classroom
William J. Furiosi II
University of Central Florida
Purpose
Lesson Description
Computer Vision Relationship
Connect computer vision with Advanced Placement Biology
instruction using computer-aided detection algorithms to improve
lung cancer diagnosis.
Introduction
The bridge between biology and computer vision is reinforced
based on the following applications within the lesson:
Overview
Increasing technological advances have led to a significant
interdisciplinary approach in the fields of biology and medicine.
Day 1: Setting the stage, students receive direct instruction in the following topics:
 General respiratory anatomy.
 Lung cancer significance, epidemiology, and diagnosis.
 Benefits and disadvantages of computed tomography (CT).
 Impact of computer vision on lung nodule detection.
Incidence of death by cause
Diseases of the heart
Cancer
Chronic lower respiratory
disease
Accidents
This lesson aims to:





Connect medicine to computer vision applications
Introduce programming principles in a public school setting
Expose high school students to genuine medical scenarios
Elaborate on cell cycling pathways in relation to cancer
Improve information synthesis, through use of multiple charts,
images, and graphs.
Education Standards
AP Curriculum Standards
3.A.2: In eukaryotes, heritable information is passed to the next
generation via processes that include the cell cycle and mitosis or
meiosis plus fertilization.
Figure 3: Comparison
of cancer deaths by
type of cancer. Lung
cancer far surpasses
any other type with
over a quarter of all
cancer-related deaths. [1]
Stroke
Brain/Nervous System
Female Breast
Colon & Rectum
Leukemia
Lung & Bronchus
Disabetes mellitus
Influenza and pneumonia
Non-Hodgkin
Lymphoma
Ovary
Kidney related disease
Pancreas
Suicide
Prostate
Practice
Day 2: Students spend time segmenting nodules via a PowerPoint tutorial and known reference.
The initial segmentation will likely
select a larger area than desired. Use
the (maximum total volume) slider to
reduce your selection to only the lung
cavity. Keep in mind the limitation of
region growing is that it may select too
much or too little of the desired area.
Figure 4: Sample slide from the student tutorial showing how to segment the lung and approximate lung
volume.
Figure 5: Sample lung cancer data showing a nodule with a maximum diameter of 48 mm
with the nodule (yellow) and lungs (salmon) segmented. Data from The Cancer Imaging
Archive.[2, 4]
Region-growing in Photoshop and image manipulation.
Facial recognition for surveillance cameras.
Population counting at large events.
Optimizing search engine results based on image characteristics.
Acknowledgements
 Salman Khokhar and Aliasghar Mortazi, UCF Ph. D. students,
for their programming instruction and troubleshooting.
 Dr. Ulas Bagci for his direction with segmentation of lung
nodules and technical expertise.
 Dr. Niels Lobo for his administration of the program,
troubleshooting, and advice as co-principal investigator.
 Dr. Mubarak Shah for his oversight as principle investigator.
Funding for this project was provided through the National Science
Foundation, grant #1542439.
Role-Playing
Day 3: Students role-play a medical exam involving symptoms typical
of lung cancer. Students are divided into one of two roles: Patient or
Medical Resident.
References
Patients and medical residents will role-play an examination,
consultation with an attending physician, and diagnosis and treatment.
Staging
T: T3
N: N0
M: M0
TNM Stage: Stage IIA
Treatment Recommendations
Surgery, likely pneumonectomy.
Chemotherapy to limit further spread.
Figure 6: Sample cancerous nodule segmented and staged
according to World Health Guidelines. The medical team
would present these findings to the patient during the
treatment stage. Data from The Cancer Imaging Archive.[2, 4]




Many thanks go to the following individuals:
Next Generation Sunshine State Standards
SC.912.L.16.8: Explain the relationship between mutation, cell
cycle, and uncontrolled cell growth potentially resulting in cancer.
Figure 1: The above diagram shows the stages in the animal cell cycle. The cell cycle is the most relevant application of this biologycomputer vision lesson. Checkpoint regulation failures, especially in the p53 regulatory pathways, can lead to unregulated cell growth
and cancer. When lung cells are exposed to toxins, like those in tobacco smoke, or radiation, they can become cancerous.
Alternative applications of computer vision are also mentioned
during instruction, including:
Liver
lzheimer's disease
Figure 2: Comparison of all deaths by cause. Cancer
is the second leading cause of death and accounts for
nearly a quarter of all deaths.[3]
Deaths by cancer type
 Use of Python to develop programs that meet research needs.
 Use of computer algorithms to detect lung nodules at sizes
smaller than manually detected by radiologists.
 Region-growing as a means of segmenting nodules and the
lungs from other structures in the body.
Tables 1 & 2: Patient history and nodule characterization references for use by
the medical team during the examination and diagnosis stages.
[1]American Cancer Society. (2014). Cancer Facts & Figures
2014. Atlanta, GA: American Cancer Society, Inc.
[2]Clark K., Vendt B., Smith K., Freymann J., Kirby J., Koppel P.,
… Prior F. (2013). The Cancer Imaging Archive (TCIA):
Maintaining and operating a public information repository.
Journal of Digital Imaging, 26(6), 1045-1057.
[3]National Center for Health Statistics. (2016). Health, United
States, 2015: With special feature on racial and ethnic health
disparities. Washington, DC: U.S. Government Printing Office.
[4]Smith K., Clark K., Bennett W., Nolan T., Kirby J., Wolfsberger
M., … Freymann J. Data from NSCLC-Radiomics-Genomics.
doi: 10.7937/K9/TCIA.2015.L4FRET6Z