This lecture will detail the creation of ultrasensitive sensors based on electronically active conjugated polymers (CPs) and carbon nanotubes (CNTs). A central concept that a single nano- or molecular-wire spanning between two electrodes would create an exceptional sensor if binding of a molecule of interest to it would block all electronic transport. The use of molecular electronic circuits to give signal gain is not limited to electrical transport and CP-based fluorescent sensors can provide ultratrace detection of chemical vapors via amplification resulting from exciton migration. Nanowire networks of CNTs provide for a practical approximation to the single nanowire scheme. These methods include abrasion deposition and selectivity is generated by covalent and/or non-covalent binding selectors/receptors to the carbon nanotubes. Sensors for a variety of materials and cross-reactive sensor arrays will be described. The use of carbon nanotube based gas sensors for the detection of ethylene and other gases relevant to agricultural and food production/storage/transportation are being specifically targeted and can be used to create systems that increase production, manage inventories, and minimize losses.
The impetus for the SENSE.nano is the recognition that novel sensors and sensing system are bound to provide previously unimaginable insight into the condition of individuals, as well as built and natural world, to positively impact people, machines, and environment. Advances in nano-sciences and nano-technologies, pursued by many at MIT, now offer unprecedented opportunities to realize designs for, and at-scale manufacturing of, unique sensors and sensing systems, while leveraging data-science and IoT infrastructure.
Understanding the brain could lead to new kinds of computational algorithms and artificial intelligences, as well as treatments for intractable disorders that affect over a billion people worldwide. However, the brain is a very complex, densely wired circuit, and understanding how it works has remained elusive. In order to map how these circuits are organized, and control their complex dynamics, we are building new tools, which include methods for physically expanding brain circuits so that we can see their building blocks, as well as molecules that make neural circuits controllable by light. Through these tools we aim to enable the systematic analysis and repair of the brain.
Polina Golland will discuss her group's research in computational analysis of MRI scans that aims to provide accurate measurements of healthy anatomy and physiology, and biomarkers of pathology. Applications range from fetal development to aging brain.
Early and accurate detection of cancer represents an enormous opportunity for sensing technologies to impact patients' lives. I will discuss several examples of diagnostic technologies developed in the Bhatia lab that employ nanosensors to detect tumors using a simple urine test for readout. This platform technology uses nanosensors to detect enzyme activity associated with cancer invasion, and generate bar-coded reporters that can be detected by multiplexed mass spectrometry or antibody-based methods such as lateral flow assays. I will close the presentation with an introduction to the Marble Center for Cancer Nanomedicine, a new growing resource for the nanomedicine community.
While trillions of sensors connected to the “Internet of Everything” (IoE) promise to transform our lives, they simultaneously pose major obstacles, which we are already encountering today. Max Shulaker presents a path towards realizing these future systems in the near-term, and shows how based on the progress of several emerging nanotechnologies (carbon nanotubes for logic, non-volatile memories for data storage, and new materials for sensing), we can begin realizing these systems today.
Trusting any data set or analysis requires a leap of faith. Beyond an acceptance of margins of error and biases, all data-driven decisions necessitate a will to believe. When it comes to data that impacts or justifies institutional decisions, this belief must exist not only in the institution's ability to be honest and rigorous with data, but in the very authority of data itself to tell us something meaningful about the world. In an era of “alternative facts” and fear-based advocacy, we must contend with this; but it may also sometimes be a symptom of data tunnel vision. How can we be better at designing the conditions for people to develop faith in our (and their) ability to do good things with data? And how can purposefully-deployed inefficiencies improve the resilience of human systems?
Advances in optics, biological sensing, medical imaging technologies, high throughput genetic sequencing is leading to massive datasets, which need to be analyzed. However, current Artificial Intelligence algorithms usually require 1000’s of examples of well-annotated datasets for high accuracy classification. Fluorescent biomarkers are important indicators of disease such as oral cancer, but imaging them can require specialized and often-expensive devices. Medical images, if diagnosed early with biomarker images and expert knowledge, can be valuable to prevent occurrences of serious systemic illnesses. In this lecture, we will discuss two convolutional neural network classifiers trained with disease signatures and fluorescent biomarker images to identify biomarkers in white light images as a per-pixel binary classification task. Once trained, the classifiers predict the location and intensity of fluorescent biomarkers in white light images without requiring specialized biomarker imaging devices or expert intervention. This generalized approach can be useful in other domains where diagnostic biomarker predicting can augment expert knowledge using standard white light images.
Medicine presents a particular problem for creating artificial intelligence (AI), because the issues and tasks involved are surprisingly subjective. Valid and useful AI requires not only reliable, unbiased, and extensive data, but also objective definitions and intentions. Assistance is most needed in day-to-day complex decision-making that requires data synthesis and integration, tasks we now approach with clinical intuition. This process is generally accepted as representing the ‘art’ of medicine despite being riddled with cognitive biases and often based on large information gaps. Resolving the subjectivity of medicine with the objectivity required for digitization—and the secondary creation of AI—first involves resolution of a number of questions: What do we want to do? What do we need to do? What can we do?
Attempts to embed machine learning-based predictive models to make products smarter, faster, cheaper, and more personalized will dominate activity in the technology industry for the foreseeable future. Veeramachaneni and MIT researchers are proposing a paradigm shift from the current practice of creating machine learning models that requires months-long discovery, exploration and “feasibility report” generation, followed by re-engineering for deployment, in favor of a rapid 8 week long process of development, understanding, validation and deployment that can executed by developers or subject matter experts using reusable APIs.