Deep Learning Evaluation for Cholangiocarcinoma Diagnosis in Bile Duct Whole Slide Images
Aadi Duggal, Vedhsai Thiriveedi, Rishitha Mantri & Rishika Singh
Lay Summary:
Our project uses advanced imaging and artificial intelligence to improve the early detection of bile duct cancer, a rare but deadly disease. By training a computer model to accurately distinguish between cancerous and non-cancerous cells, we aim to enhance diagnosis and help doctors make better treatment decisions, potentially saving lives.
Abstract:
Cholangiocarcinoma, commonly known as bile duct cancer, is a rare but deadly disease, typically diagnosed at advanced stages due to its subtle early symptoms. Given the similarity in symptoms between cholangiocarcinoma and other benign conditions, traditional diagnostic methods often struggle with classification, resulting in frequent misdiagnosis. With the overall survival rate of this disease being ~10%, there is an urgent need for more precise and accurate diagnostic approaches. In this study, we aim to improve early detection and classification of bile duct cancer through advanced imaging and deep learning techniques. Whole Slide Images (WSI) were utilized to enhance the accuracy of cell classification within bile duct tissues with machine learning methods. Whole slide images were annotated using the Segment Annotation Model (SAM) for various cluster types, including atypical, benign, suspicious positive (SUSPos), and white blood cells (WBCs). These were then used as inputs for YOLO (You Only Look Once) object detection to identify the individual cells. A Convolutional Neural Network (CNN) was then used based on the AlexNet architecture and was trained for binary classification of atypical and benign cells. The CNN showed strong performance with an AUC of 0.94 and 88% accuracy on the test dataset, demonstrating its effectiveness in classification. These findings demonstrate the ability of deep learning to advance the diagnosis of bile duct cancer, thereby improving clinical outcomes.
Q&A:
Bios: Aadi Duggal,Vedhsai Thiriveedi,Rishitha Mantri,Rishika Singh
Program Track: Advanced Research
GitHub Username:
aadiduggal -Aadi Duggal
Vedhsai-codes -Vedhsai Thiriveedi
rishiiimantri -Rishitha Mantri
rishika-singhh -Rishika Singh
What was your favorite seminar? Why?
My favorite seminar was the second seminar of the internship featuring Jason Wei and Ryan Urbanowicz. I found this this seminar really cool since Jason works at Open AI and helped with the development of chat gpt. Over the past 2ish years chat GPT has blown up in popularity and can do so much cool stuff like writing essays, answering extremely hard questions, and having a lot of accuracy. I have personally used chatGPT and it was amazing to get to listen to and ask questions to someone who works on it and can give us updates and credible information about gpt and whats happening next. A second reason why I loved this seminar was because of Ryan and his expertise on machine learning. I have been fascinated with machine learning for about a year and was able to do a research project focusing on machine learning before this internship, so hearing from someone who uses machine learning for a different area that I have little knowledge of was super awesome since I could expand my own machine learning knowledge. -Aadi Duggal
My favorite seminar was by Alex Xu and his seminar on spatial biology technologies. I found that I actually enjoyed the subject and learned alot. -Vedhsai Thiriveedi
The seminar on multimodel analysis was fascinating to me because it was really nice to see how I could make sense of qualitative data by looking at so many different aspects like spatial, visual, etc. -Rishitha Mantri
My favorite seminar was done by Jason Wei because I loved hearing his unique experience using machine learning and even how he works at OpenAI and making a tool so many use today. -Rishika Singh
If you were to summarize your summer internship experience in one sentence, what would it be?
A journey of learning, eustress, and growth where I applied my skills to real-world challenges and gained invaluable insights into my field. -Aadi Duggal
An amazing experience where you get to learn about medicine and machine learning with a team. -Vedhsai Thiriveedi
From this summer internship, I learned about the impact machine learning could have in clinical settings and had the opportunity to make an impact myself. -Rishitha Mantri
A unique experience allowing me to grow, learn, and a valuable asset to me in the future. -Rishika Singh