Upcoming Events
Currently there are no events.
Past Events
Enhancing Interpretability in DL Models for Breast Cancer Detection
(11:00 am)
Title: Enhancing Interpretability in DL Models for Breast Cancer Detection
Abstract: Breast cancer is one of the most common cancers found primarily in women, affecting 1 in 8 women in Canada over their lifetime, according to the Public Health Agency of Canada. The early diagnosis of breast cancer can reduce the spread of cancer cells and the risk of death. Machine learning (ML) systems can help healthcare professionals diagnose breast cancer with better accuracy by seeing the underlying patterns in the images that the radiologist might find difficult to detect early on. While deep learning (DL) models can obtain high accuracies, their black-box nature inhibits their ability to explain how they came to their conclusion and their regions of interest (ROI), creating distrust between DL systems and the radiologist. This research employs two CNN model architectures - AlexNet and Houby & Yassin's CNN – to classify recently obtained mammographic images from the KAU-BCMD dataset into six BI-RAD categories (0 - 6). To resolve the problem, SHAP, an explainability approach based on Shapley values, has been used to assess the role that each feature has towards the prediction process. There were two SHAP variants that were applied: PartitionSHAP and DeepSHAP. The method of employing the dataset through the models is by 1) preprocessing the dataset, 2) classifying the dataset in the BI-RADs 0-6, and 3) obtaining the SHAP plots identifying the model's ROIs. The proposed methods resulted in high-performance metrics for AlexNet and Houby & Yassin's CNN with an AUC and F1-score of 0.983 and 0.888, and 0.982 and 0.891, respectively.
A Comparison of Autoencoders and Variational Autoencoders for Anomaly Detection in Dermoscopic Images
(11:00 am)
Title: A Comparison of Autoencoders and Variational Autoencoders for Anomaly Detection in Dermoscopic Images
Abstract:
The early detection and diagnoses of skin abnormalities are crucial for effective treatment and management of skin diseases. This paper explores the application of deep learning techniques for skin tissue analysis, focusing on the detection of abnormalities from dermoscopic images. Unsupervised learning methods like Autoencoders (AE) and Variational Autoencoders (VAE) save resources by eliminating the need for labeled data, making them more efficient and scalable than supervised learning. We compare the performance of AE and VAE architectures in developing a robust model capable of distinguishing between benign and malignant skin lesions.
This study uses the HAM10000 dataset of 10,015 dermoscopic images, divided into seven classes representing both benign and malignant diagnostic categories. The dataset was split into benign (normal) and malignant (anomalous) cases. The models were trained to learn features of the normal data and generate reconstructions of these images. The reconstruction error measures how accurately the model interprets the image features. An optimal decision boundary is chosen to classify images as benign or malignant based on their reconstruction error.
The AE and VAE models were evaluated using accuracy, F1-score, and False Negative Rate (FNR). Minimizing FNR is crucial in healthcare as it indicates missed malignant cases. Experimental results, averaged over 30 training runs, demonstrate that the VAE outperforms the AE in accuracy (69.31% vs. 68.80%), while the AE surpasses the VAE with a higher F1-score (70.30% vs. 69.45) and a lower FNR (26.15% vs. 30.19%). These findings suggest the AE architecture is promising for automated skin cancer detection, potentially leading to more accurate and timely diagnoses and improved patient outcomes.
Exploring the Intersection of AI and Entrepreneurship: Lessons and Insights from Silicon Valley to Acadia University
(3:00 pm - 4:00 pm)
Talk by Dr. Junwei Zhang
Tech Lead Engineer/Investment Director
Date: March 8th
Time: 3:00PM - 4:00PM
Location: Online
Teams link: Click here to join the meeting
Email: junweizhang23@gmail.com
Abstract: In today's fast-paced digital world, Artificial Intelligence (AI) is transforming industries and driving the growth of startups. Junwei Zhang, with experience at Uber, Microsoft, and DoorDash, will share how Al can be integrated into new businesses, highlighting both challenges and opportunities.
With a background in computer vision and parallel computing, and hands-on leadership in tech projects, Zhang will discuss his journey through the dynamic tech and investment scene in Silicon Valley. The talk will focus on the latest developments in Al and the essentials of startup growth and venture capital, all based on his real-world experiences.
This session is intended as a mutual exchange of ideas, offering practical insights into how technology and entrepreneurship come together. It aims to inspire students with a clear understanding of the Al field and provide actionable advice for those interested in starting their own ventures or exploring the world of investment. Join us to learn how Al and entrepreneurship can shape the future and drive innovation.
Brief Bio: I am Junwei Zhang, and I am both an engineer and a venture capital investor with a robust background in technology and research. I earned my Ph.D. in Applied Mathematics and Statistics from Stony Brook University in New York. My journey in the tech industry began prior to the Uber IPO when I joined Uber as a software engineer. Following this, I embarked on roles within the Microsoft Azure Cloud + AI team and the DoorDash Growth team, contributing to significant projects and innovations. In addition to my industry experience, I serve as an associate editor for the IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), where I oversee contributions in areas pivotal to our field. My academic contributions include over 10 papers published in top-tier conferences and journals such as ICCV, CAD, CAGD, and TPDS, covering a range of topics from computer vision to geometry processing and parallel computing.
Venture capital investment is another avenue through which I engage with the tech ecosystem. I have led investments in startups, including those that empower Stanford researchers to leverage new AI technologies for mental health initiatives. Moreover, I am an active board member of a Silicon Valley-based tech entrepreneurship leadership community, where we organize monthly webinar discussions to foster innovation and leadership in the tech sector.
CS Seminar: Exploring the Intersection of AI and Entrepreneurship: Lessons and Insights from Silicon Valley to Acadia University (Seminar held with the LaunchBox)
(2:30 pm - 3:30 pm)
Talk by Dr. Junwei Zhang
Tech Lead Engineer/Investment Director
Date: March 8th
Time: 2:30pm – 3:30pm
Location: Online
Teams link: Click here to join the meeting
Profile: LinkedIn
Email: junweizhang23@gmail.com
Abstract: In our rapidly evolving digital age, the impact of Artificial Intelligence on industries and the dynamism of startup ecosystems offer a unique canvas for innovation. Junwei Zhang, who has contributed to technology and venture capital through roles at companies including Uber, Microsoft, and DoorDash, shares insights into the integration of AI with entrepreneurial ventures, reflecting on the challenges and opportunities this convergence presents.
Drawing from a solid foundation in computer vision and parallel computing, alongside practical involvement in tech project leadership, Zhang aims to share his pathway and insights gained from navigating the vibrant tech and investment landscape of Silicon Valley. The discussion will allocate a thoughtful portion to AI — addressing its current frontiers and challenges — while also delving into the realms of startup development and venture capital, informed by Zhang's direct experiences and collaborative endeavors.
This session is designed to be an exchange of learnings rather than a highlight of personal accolades. It seeks to offer a nuanced understanding of how technology and entrepreneurship intersect, providing valuable perspectives for those interested in shaping the future of tech and business. Attendees will gain a broader understanding of the AI landscape, alongside practical advice on entrepreneurship and investment, fostering a spirit of innovation and growth.
Join us for a journey through the complexities and triumphs of merging AI with entrepreneurship, guided by the shared experiences from Silicon Valley's ecosystem to the academic and entrepreneurial communities at Acadia University.
Brief Bio: I am Junwei Zhang, and I am both an engineer and a venture capital investor with a robust background in technology and research. I earned my Ph.D. in Applied Mathematics and Statistics from Stony Brook University in New York. My journey in the tech industry began prior to the Uber IPO when I joined Uber as a software engineer. Following this, I embarked on roles within the Microsoft Azure Cloud + AI team and the DoorDash Growth team, contributing to significant projects and innovations. In addition to my industry experience, I serve as an associate editor for the IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), where I oversee contributions in areas pivotal to our field. My academic contributions include over 10 papers published in top-tier conferences and journals such as ICCV, CAD, CAGD, and TPDS, covering a range of topics from computer vision to geometry processing and parallel computing.
Venture capital investment is another avenue through which I engage with the tech ecosystem. I have led investments in startups, including those that empower Stanford researchers to leverage new AI technologies for mental health initiatives. Moreover, I am an active board member of a Silicon Valley-based tech entrepreneurship leadership community, where we organize monthly webinar discussions to foster innovation and leadership in the tech sector.
Poster Link
CS seminar: AI for healthcare: CNNs, Transformers, and Beyond
Title: AI for healthcare - CNNs, Transformers, and Beyond.
Presenter: Moulay Akhloufi, Professor at University of Moncton and Founder of the Perception, Robotics and Intelligent Machines (PRIME) research group
Summary: In recent years, we have seen important progress in the field of AI for healthcare, largely attributed to the impressive progress resulting from the application of deep learning. Various medical disciplines, including ophthalmology and radiology, have experienced the positive impact of these advancements. This talk will present the latest developments in leveraging deep learning for medical imaging. This includes the use of Convolutional Neural Networks (CNNs) and deep Transformers. Additionally, I will illustrate how the integration of deep ensemble learning improves the performance of specific tasks in this domain. I will also explore the application of deep learning in detecting eye diseases like diabetic retinopathy and AMD, among others. Moreover, I will discuss the utilization of deep learning and transformers in radiology for the identification and diagnosis of a diverse type of diseases. In addition, cases in oncology, particularly in breast cancer and skin cancer detection, will be showcased. Given the importance of understanding the decisions made by these algorithms, I will provide examples of explainability techniques. Furthermore, I will emphasize the significance of federated learning in enhancing the training performance of deep models while ensuring privacy. Finally, I will touch upon potential research directions in this evolving field.
Short bio: Professor Moulay Akhloufi holds a Bachelor of Science in Physics from the University Abdelmalek Essaadi (Morocco) and a Bachelor of Engineering from Telecom Saint-Etienne (France). He has a Master's and Ph.D. in Electrical Engineering from Ecole Polytechnique of Montreal and Laval University (Canada), respectively. Additionally, he holds an MBA from Laval University. Presently, Professor Akhloufi serves as Professor in Computer Science at Université de Moncton, where he leads the Perception, Robotics, and Intelligent Machines (PRIME) research lab, and holds the position of Director at the Center for Artificial Intelligence NB Power. Prior to joining Université de Moncton in 2016, he gained valuable experience in the industry and in technology transfer within the fields of machine vision and robotics. Professor Akhloufi's research expertise spans across the domains of artificial intelligence, computer vision, and intelligent robotic systems, where he has contributed to over two hundred publications. Additionally, he is a Senior Member of the Institute of Electrical and Electronics Engineers (IEEE), and a member of the Society of Photo-Optical Instrumentation Engineers (SPIE).
