Using a Machine Learning Algorithm to Detect Basal Cell Carcinoma in Microscope Slides of Mohs Excisions
School Name
Spring Valley High School
Grade Level
11th Grade
Presentation Topic
Computer Science
Presentation Type
Non-Mentored
Written Paper Award
1st Place
Abstract
It has been estimated that approximately 20% of Americans will develop some form of skin cancer in their lifetime; over 80% of these cases will be basal cell carcinoma. The process of detecting and locating basal cell carcinoma in microscope slides of Mohs excisions is time-consuming for Mohs surgeons, and operations can take hours to complete. Machine learning is a new field of artificial intelligence that has started to see growth in the field of medicine over the past few years. This experiment was conducted in order to study how accurately a supervised machine learning program would be able to detect the presence of basal cell carcinoma in microscope slides of Mohs excisions. It was hypothesized that the algorithm, after training, would be able to detect examples of basal cell carcinoma at an accuracy rate of 95% or higher. There were 1660 slides in total, with 1490 being used for the training period and the final 170 for testing. The final model was trained for 5 hours and had an average precision of 0.995. The precision and recall values were both 0.976 at a score threshold of 0.5.
Recommended Citation
Zhang, Luke, "Using a Machine Learning Algorithm to Detect Basal Cell Carcinoma in Microscope Slides of Mohs Excisions" (2019). South Carolina Junior Academy of Science. 294.
https://scholarexchange.furman.edu/scjas/2019/all/294
Location
Founders Hall 140 B
Start Date
3-30-2019 9:45 AM
Presentation Format
Oral and Written
Group Project
No
Using a Machine Learning Algorithm to Detect Basal Cell Carcinoma in Microscope Slides of Mohs Excisions
Founders Hall 140 B
It has been estimated that approximately 20% of Americans will develop some form of skin cancer in their lifetime; over 80% of these cases will be basal cell carcinoma. The process of detecting and locating basal cell carcinoma in microscope slides of Mohs excisions is time-consuming for Mohs surgeons, and operations can take hours to complete. Machine learning is a new field of artificial intelligence that has started to see growth in the field of medicine over the past few years. This experiment was conducted in order to study how accurately a supervised machine learning program would be able to detect the presence of basal cell carcinoma in microscope slides of Mohs excisions. It was hypothesized that the algorithm, after training, would be able to detect examples of basal cell carcinoma at an accuracy rate of 95% or higher. There were 1660 slides in total, with 1490 being used for the training period and the final 170 for testing. The final model was trained for 5 hours and had an average precision of 0.995. The precision and recall values were both 0.976 at a score threshold of 0.5.
