AI & Deep Learning Laboratory
With the rise of neural networks, computer systems that mimic the neuronal structure of the brain, there is a great potential for machine-assisted interpretation in a clinical context. Using the latest advancements in Deep Learning, a machine learning technique well suited for image analysis, the AI and deep learning researchers in the Department have been able train models to classify medical images.
The lab welcomes collaborations with other faculty members, as well as medical students and resident trainees who want to work at the cutting edge of applied AI research.
What is Deep Learning?
Deep learning is a branch of artificial intelligence that uses artificial neural networks. Just like the cortex of our brain, the networks are arranged in multiple layers, and can represent information at multiple levels.
Machine learning is particularly useful because computers are not explicitly programmed. In the past, researchers would have to code multiple rules to recognize a basic part of the image. The sheer amount of variation between images makes it a daunting task to produce such computer-aided detection systems. However, now the machine needs to be trained on a sufficient number of examples.
Projects
Images taken from Lee, Honglak, et al. "Convolutional deep belief networks for scalable unsupervised learning of hierarchical representations."
Along with faculty and resident collaborators, a number of projects have been completed that provide a proof of concept for the utility of neural networks in the clinical workflow.
Areas of research in the lab include identification of pneumothoraces on chest radiography, feeding tube placement, acute brain hemorrhages on head CT and bone fractures, as well as image segmentation and numerous other quantitative projects involving deep learning.
Selected Publications
Lakhani P., Sundaram B. "Deep Learning on Chest Radiography: Automated Classification of Pulmonary Tuberculosis Using Convolutional Neural Networks." Radiology, In Press.
Lakhani P., Flanders A. “Deep Convolutional Neural Networks for Endotracheal tube position and X-ray Image Classification: Challenges and Opportunities.” Conference of Machine Intelligence in Medical Imaging, September 2016.
Desai V, Lakhani P, Flanders A. “Application of Deep Learning in Neuroradiology: Automated Detection of Basal Ganglia Hemorrhage” American Society of Neuoradiology Annual Meeting. April 2017.