Center for Artificial Intelligence, Innovation Research & Equity (CAIIRE)

Our team is developing algorithms that use biometric data from personal wearable devices to predict postoperative complications after major operations. Currently, there is a major gap in patient monitoring in the weeks following hospital discharge – a time when patients remain at high-risk for post-operative complications. As we move into a new era in which wearable devices are becoming ubiquitous, wearable tools that proactively alert patients and providers about complications can play a transformative role in improving patient outcomes.

While lung cancer screening with a low-dose computed tomography (CT) scan is the best way to detect lung cancer early, many high-risk patients remain ineligible for lung cancer screening. Notably, there are major racial and sex-based disparities in lung cancer screening eligibility. A team led by CAIIRE faculty members Dr. Lecia Sequist, Dr. Florian Fintelmann, and MIT researcher Dr. Regina Barzilay developed Sybil, an AI tool that can improve early lung cancer detection. Sybil is a 3-D convolutional neural network that examines an entire chest CT scan and estimates probabilities that lung cancer will be diagnosed in each of the next six years. In line with our center’s equity focus, the Sybil team is training the model on a scans from a diverse patient population. Notably, the model will be trained on scans from our team’s INSPIRE study, a prospective lung cancer screening study of Black women undergoing lung cancer screening.

We are developing computer vision algorithms to automatically analyze and interpret videos of operations as they are occurring. Our goal is to train AI to understand what is happening during operations, in real-time. In the future, we strive to develop computer-vision based tools to support intra-operative decisions and identify patients who are at high-risk for postoperative complications. Additionally, we hope to develop computer-vision based tools to improve education of surgical trainees.

Our team is developing strategies to combine virtual reality (VR) and olfactory stimulation to improve patient sleep, pain, anxiety, and sleep quality while in the hospital. Patients in our pilot study are offered an intervention consisting of an immersive audiovisual VR experience displaying jellyfish pulsing at 40 beats per minute (half of that average heart rate) accompanied by aquatic white noise and heart sounds. The VR experience is paired with a lavender olfactory stimulus (OS). At night, patients are provided with a sleeping mask infused with the same OS. We hypothesize, based on preliminary research from collaborator Dr. Judith Amores, that the overnight OS may trigger memory reactivation of the relaxing VR experience, and resultantly reduce pain, lower anxiety, and improve sleep quality.

Spread through air spaces (STAS) is a unique phenomenon in lung cancer that was recently recognized by the WHO in 2015. It is believed that invasion of airspace may be associated with worse recurrence-free and overall survival, but there has so far been no reliable way of identifying potential STAS in the pre-operative setting. We are working to identify clinical and radiomic factors associated with STAS and develop a combined computer-vision based clinical radiomic model to predict STAS status pre-operatively.