« The Sylvester Comprehensive Cancer Center of the University of Miami named 71st NCI designated Cancer Center

Broad@15 – In 2004, the Broad Institute of MIT and Harvard launched with a mission to improve human health »

MinneBOS 2019 Field Guide to Data Science & Emerging Tech in the Boston Community, August 22, 2019 8AM to 5PM at Boston University Questrom School of Business, 595 Commonwealth Avenue, Boston, MA

July 31, 2019 by 2012pharmaceutical

MinneBOS 2019, Field Guide to Data Science & Emerging Tech in the Boston Community

August 22, 2019, 8AM to 5PM at Boston University Questrom School of Business, 595 Commonwealth Avenue, Boston, MA

MinneBOS – Boston’s Field Guide to Data Science & Emerging Tech

https://minnebos.sched.com/list/descriptions/

Announcement

Leaders in Pharmaceutical Business Intelligence (LPBI) Group

REAL TIME Press Coverage for

http://pharmaceuticalintelligence.com

Aviva Lev-Ari, PhD, RN

Director & Founder, Leaders in Pharmaceutical Business Intelligence (LPBI) Group, Boston

Editor-in-Chief, Open Access Online Scientific Journal, http://pharmaceuticalintelligence.com

Editor-in-Chief, BioMed e-Series, 16 Volumes in Medicine, https://pharmaceuticalintelligence.com/biomed-e-books/

#MinneBos

Logo, Leaders in Pharmaceutical Business Intelligence (LPBI) Group, Boston

Our BioMed e-series

WE ARE ON AMAZON.COM

ekWGNqA

UPDATED AGENDA

Advancing Cancer Research with Deep Learning Image Analysis

Histopathological images are the gold standard tool for cancer diagnosis, whose interpretation requires manual inspection by expert pathologists. This process is time-consuming for the patients and subject to human error. Recent advances in deep learning models, particularly convolutional neural networks, combined with big databases of patient histopathology images will pave the path for cancer researchers to create more accurate guiding tools for pathologists. In this talk, I will review the latest advances of big data in healthcare analytics and focus on deep learning applications in cancer research. Targeted at a general audience, I will provide a high-level overview of technical concepts in deep learning image analysis, and describe a typical cloud-based workflow for tackling such big data problems. I will conclude my talk by sharing some of our most recent results based on a wide range of cancer types.

Speakers

Mohammad Soltanieh-ha, PhD

Clinical Assistant Professor, Boston University – Questrom

Mohammad is a faculty at Boston University, Questrom School of Business, where he teaches data analytics and big data to master’s students. Mohammad’s current research area involves deep learning and its applications in cancer research.

10:15am

Break

10:30am

Deep learning image recognition and classification models for fashion items

Large scale image recognition and classification is an interesting and challenging problem. This case study uses fashion-MNIST dataset that involves 60000 training images and 10000 testing images. Several popular deep learning models are explored in this study to arrive at a suitable model with high accuracy. Although convolutional neural networks have emerged as a gold-standard for image recognition and classification problems due to speed and accuracy advantages, arriving at an optimal model and making several choices at the time of specifying model architecture, is still a challenging task. This case study provides the best practices and interesting insights.

Speakers

Bharatendra Rai

Professor, UMass Dartmouth

Bharatendra Rai, Ph.D. is Professor of Business Analytics in the Charlton College of Business at UMass Dartmouth. His research interests include machine learning & deep learning applications.

Train data: 60,000
Test data: 10,000
Dataset available from Google MNIST Fashion Data – items in DB: data already labelled
Label and Description
Architecture: Input >> Conv >> Conv >> Pooling >> Dropout << Dense <<Flatten << Dropout >> Output
CNN vs Fully connected: 320 parameters: 3x3x1x32 + [32 BIAS TERM] = 320 vs
fully connected network parameters is 16 million
Train the model: 15 iterations – Training and Validation
Actual vs Predicted: 94% was classified correctly = Accuracy: 94% 5974 vs 4700 (78%)
Confusion Matrix – Test 720 correctly classified for item 6 – Probability va Actual Vs Predicted
Image generation: Noise . gnerator Network > fake Image vs Real image – GAN Loss va Discriminator Loss
CNN network help reduce # of parameter
Droppot layers can help reduce overfitting
validation split of x%chooses last x% of train data
Generation of new data is challenging

11:00am

Break

11:15am

Rapid Data Science

Most companies today require fast, traceable, and actionable answers to their data questions. This talk will present the structure of the data science process along with cutting edge developments in computing and data science technology (DST) with direct applications to real world problems (with a lot of pictures!). Everything from modeling to team building will be discussed, with clear business applications.

Speakers

Erez Kaminski

Leaders Global Operations Fellow, MIT

Erez has spent his career helping companies solve problems using data science. He is currently a graduate student in computer science and business at MIT. Previously, he worked in data science at Amgen Inc. and as a technologist at Wolfram Research.

12:00pm

Lunch

1:00pm

Health and Healthcare Data Visualization – See how you’re doing

Health and healthcare organizations are swimming in data but few have the skills to show and see the story in their data using the best practices of data visualization. This presentation raises awareness about the research that inform these best practice and stories from the front of groups who are embracing them and re-imagining how they display their data and information. These groups include the NYC Dept of Health & Mental Hygiene, The Centers for Medicare and Medicaid (CMS), and leading medical centers and providers across the country.

Speakers

Katherine Rowell

Co-Founder & Principal, Health Data Viz

Katherine Rowell is a health, healthcare, and data visualization expert. She is Co-founder and Principal of HealthDataViz, a Boston firm that specializes in helping healthcare organizations organize, design and present visual displays of data to inform their decisions and stimulate… Read More →

dashboard for Hospital CEOs

1:45pm

Break

2:00pm

AI in Healthcare

Benefits, challenges and impact of AI and Cybersecurity on medicine.

Speakers

Vinit Nijhawan

Lecturer, Boston University

Vinit Nijhawan is an Entrepreneur, Academic, and Board Member with a track record of success, including 4 startups in 20 years.

US: Spends the most on Health Care (HC) death per 100K people is the highest
Eric Topol – Diagnosis is not done correctly, AI will help with diagnosis
Diagnosis — AI will have the most impact; VIRAL infections are diagnosed as bacterial infections and get antibiotics for treatment
Image Classification my ML – decline below to human misclassification
Training Data sets – Big data
Algorithms getting better
Data Capture getting better – HC as well
Investment in HC is the greatest
SECURITY related to Implentable Medical Devices = security attacks – hacking and sending signal to implentable devices

2:45pm

Break

3:00pm

Empower

Patient centric AI: Saving lives with ML driven hospital interventions

This presentation will cover the use of machine learning for maximizing the impact of a hospital readmissions intervention program. With machine learning, clinical care teams can identify and focus their intervention efforts on patients with the highest risk of readmission. The talk will go over the goals, logistics, and considerations for defining, implementing, and measuring our ML driven intervention program. While covering some technical details, this presentation will focus on the business implementation of advanced technology for helping people live healthier lives.

Speakers

Miguel Martinez

Data Scientist, Optum

Miguel Martinez is a Data Scientist at Optum Enterprise Analytics. Relied on as a tech lead in advancing AI healthcare initiatives, he is passionate about identifying and developing data science solutions for the benefit of organizations and people.

3:30pm

Break

3:45pm

Using Ontologies to Power AI Systems

There’s a great deal of confusion about the role of a knowledge architecture in artificial intelligence projects. Some people don’t believe that any reference data is necessary. But in reality reference data is required- even if there is no metadata or architecture definitions outside defined externally for an AI algorithm, someone has made the decisions about architecture and classification within the program. However, this will not work for every organization because there are terms, workflows, product attributes, and organizing principles that are unique to the organization and that need to be defined for AI tools to work most effectively.

Speakers

Seth Earley

CEO, Earley Information Science

Seth Earley is a published author and public speaker about artificial intelligence and information architecture. He wrote “There’s no AI without IA” which has become an industry catchphrase used by a number of people including Ginny Rometty, the CEO of IBM.

Ontology, taxonomies, thesauri – conceptual relationships
Object-Oriented Programming and Information Architecture using AI is Old wine in new bottles

4:15pm

Networking Social in Atrium

TBA

Senior Leadership Panel: Future Directions of Analytics

This panel includes senior leaders from across industry, academia & government to discuss challenges they are tackling, needs they anticipate and goals they will achieve

Moderators