Improved lung cancer screening with cognitive computing
Principal Investigator
Name
Shawn Stapleton
Degrees
PhD
Institution
Philips Research North America
Position Title
Senior Data Scientist
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-241
Initial CDAS Request Approval
Sep 16, 2016
Title
Improved lung cancer screening with cognitive computing
Summary
Every year, 200,000 people are diagnosed with lung cancer resulting in 160,000 deaths per year (450 people per day) in the US alone. Early detection can increase the five-year survival rate for stage 1 lung cancer to nearly 90%. Screening patients with low-dose computed tomography (LDCT) has been proven to find lung cancer at its earliest stages ad reduce mortality. As a result, the number of lung cancer screenings by LDCT is expected to increase rapidly.
To address this new paradigm, we plan to leverage rich, high level features with deep learning and context with joint modeling such that clinicians and hospitals can provide optimal lung cancer screening both in terms of patient outcomes and clinical workflow. To achieve this goal we plan to leverage algorithms built from the largescale, multi-site observational NLST dataset. With this rich information, we test the feasibility of improving the differentiation of an individual patient’s risk, assisting clinicians in directing patients to optimal clinical care, and providing that information in an automated fashion to fits seamlessly in clinical workflow.
To address this new paradigm, we plan to leverage rich, high level features with deep learning and context with joint modeling such that clinicians and hospitals can provide optimal lung cancer screening both in terms of patient outcomes and clinical workflow. To achieve this goal we plan to leverage algorithms built from the largescale, multi-site observational NLST dataset. With this rich information, we test the feasibility of improving the differentiation of an individual patient’s risk, assisting clinicians in directing patients to optimal clinical care, and providing that information in an automated fashion to fits seamlessly in clinical workflow.
Aims
Aim 1: Develop Automated algorithms to reduce time and cognitive load for radiologists
Aim 2: Improve predictive strength such that unnecessary escalation and resultant complications due to false positives reads and overdiagnosis are prevented.
Aim 3: Develop algorithms that can handle high dimensionality data from radiology and digital pathology and provide improved prognostic indicators of malignancy.
Collaborators
Teun Heuvel, Philips Research Eindhoven
Arkadiusz Sitek, Philips Research North America
Tobias Klinder, Philips Research Hamburg
Rafael Wiemker, Philips Research Hamburg
Amir Tahmasebi, Philips Research North America
Sandeep Dalal, Philips Research North America
Related Publications
-
Towards radiologist-level cancer risk assessment in CT lung screening using deep learning.
Trajanovski S, Mavroeidis D, Swisher CL, Gebre BG, Veeling BS, Wiemker R, Klinder T, Tahmasebi A, Regis SM, Wald C, McKee BJ, Flacke S, MacMahon H, Pien H
Comput Med Imaging Graph. 2021 Jun; Volume 90: Pages 101883 PUBMED -
Lung-RADS Version 1.0 versus Lung-RADS Version 1.1: Comparison of Categories Using Nodules from the National Lung Screening Trial.
Kastner J, Hossain R, Jeudy J, Dako F, Mehta V, Dalal S, Dharaiya E, White C
Radiology. 2021 May 4; Pages 203704 PUBMED -
Vancouver Risk Calculator Compared with ACR Lung-RADS in Predicting Malignancy: Analysis of the National Lung Screening Trial.
White CS, Dharaiya E, Dalal S, Chen R, Haramati LB
Radiology. 2019 Apr; Volume 291 (Issue 1): Pages 205-211 PUBMED -
The Vancouver Lung Cancer Risk Prediction Model: Assessment by Using a Subset of the National Lung Screening Trial Cohort.
White CS, Dharaiya E, Campbell E, Boroczky L
Radiology. 2016 Oct; Pages 152627 PUBMED