Project Data Sphere’s Images & Algorithms (I&A) Program presented preliminary results from its autoRECIST project at the FDA Research Grand Rounds on Feb. 25. Nearly 200 attendees participated in this webinar which generated a great discussion.
The goal of the autoRECIST project is to develop deep-learning algorithms to reduce the time and cost and improve the performance of imaging in clinical trials, shortening the time from discovery to implementation, improving the accuracy of the reviews, and, ultimately, improving patient lives.
In oncology clinical trials, the overall assessment of tumor burden and response to therapy is estimated by a set of complex quantitative and qualitative criteria called Response Evaluation Criteria in Solid Tumors (RECIST). To perform a RECIST assessment, a radiologist reads Computed Tomography Digital Imaging and Communications in Medicine format (CT DICOM) images, identifies measurable lesions, picks two target lesions per organ (in up to five organs per patient), and records the largest diameter in target lesions. The same tumor images and measurements are then evaluated by an independent radiologist. An average of 30% discordance in radiological interpretation has been reported between readers.
Dr. Asba (AT) Tasneem, PDS Executive Director of the program, provided an overview of the work and discussed a four-year roadmap for the program. The two Principal Investigators at Columbia University Medical Center — Dr. Binsheng Zhao, Director, Computational Image Analysis Lab, Department of Radiology; and Dr. Larry Schwartz, Chairman, Department of Radiology — presented results from developing Liver Artificial Intelligence (AI) – the foundational AI which detects and segments liver lesions.
In the next four years, the autoRECIST project will develop deep learning algorithms to 1) calculate RECIST assessment based on volumetrics measurements of all lesions (autoVOL); and 2) automate the current RECIST 1.1 (autoRECIST).
For more information on the autoRECIST project please contact Asba (AT) Tasneem, PhD, Executive Director, Images and Algorithms Program ([email protected]).
The goal of the autoRECIST project is to develop deep-learning algorithms to reduce the time and cost and improve the performance of imaging in clinical trials, shortening the time from discovery to implementation, improving the accuracy of the reviews, and, ultimately, improving patient lives.
In oncology clinical trials, the overall assessment of tumor burden and response to therapy is estimated by a set of complex quantitative and qualitative criteria called Response Evaluation Criteria in Solid Tumors (RECIST). To perform a RECIST assessment, a radiologist reads Computed Tomography Digital Imaging and Communications in Medicine format (CT DICOM) images, identifies measurable lesions, picks two target lesions per organ (in up to five organs per patient), and records the largest diameter in target lesions. The same tumor images and measurements are then evaluated by an independent radiologist. An average of 30% discordance in radiological interpretation has been reported between readers.
Dr. Asba (AT) Tasneem, PDS Executive Director of the program, provided an overview of the work and discussed a four-year roadmap for the program. The two Principal Investigators at Columbia University Medical Center — Dr. Binsheng Zhao, Director, Computational Image Analysis Lab, Department of Radiology; and Dr. Larry Schwartz, Chairman, Department of Radiology — presented results from developing Liver Artificial Intelligence (AI) – the foundational AI which detects and segments liver lesions.
In the next four years, the autoRECIST project will develop deep learning algorithms to 1) calculate RECIST assessment based on volumetrics measurements of all lesions (autoVOL); and 2) automate the current RECIST 1.1 (autoRECIST).
For more information on the autoRECIST project please contact Asba (AT) Tasneem, PhD, Executive Director, Images and Algorithms Program ([email protected]).