Imaging Prostate Cancer
Posted: Nov 01, 2018
POSTED: May 30, 2018
In June, Prostatepedia is talking about collaborations between health care and tech for prostate cancer.
Dr. Charles Snuffy Myers frames the conversations for us.
Prostapedia’s home office is in the middle of Silicon Valley, which is a great place to be if you want to learn about the application of cutting edge technology to prostate cancer. This issue’s focus is on collaborations between clinical researchers and those with applicable technology expertise. The topics range from genome sequencing to telemedicine to phone app development.
Several things stand out. First, prostate cancer has left the era of the solitary genius and entered an era where teams of investigators with diverse skill sets are doing the most exciting work. This is necessary because of the explosion in laboratory science and computer techniques for analyzing large data sets—i.e., big data. Full time clinicians do not have time to keep up with the state of the art in these fields. However, their clinical experience is important in defining the nature of the clinical problems that need to be addressed. Additionally, the clinicians design and execute the clinical studies.
The second major development has been the recent use of machine learning to analyze data. Machine learning, especially deep neural networks, excel at recognizing patterns in large data sets and images. Already, machine learning has had success in reading radiologic images and has matched skilled dermatologists in detecting skin cancers.
This approach has great promise as a means of detecting associations between complex genomic data and clinical outcomes. To give you a sense of the scope of the big data problem we face, comprehensive genome sequencing can easily yield a terabyte of data per patient: this is equivalent to 2,000 hours of music on a CD. Imagine looking for patterns associated with clinical outcome in hundreds or thousands of patients? Machine learning involves several distinct steps. In the first step, the neural network is trained to recognize patterns. In the second step, the trained network’s performance is evaluated on a second data set. In the third step, the network is used in an ongoing manner to solve problems. The first step can be computationally intensive and with today’s technology typically requires expensive hardware. However, once the network has been trained, the actual use of the neural network in problem solving is much less demanding in terms of computer hardware.
Telemedicine represents another potential major advance. In this issue, Dr. Matthew Galsky does an excellent job outlining how telemedicine might improve the conduct of clinical trials. As he points out, many patients live a considerable distance away from clinicians doing clinical trials and this is a factor that limits patient accrual to clinical trials. Telemedicine has the potential to reduce the number of trips a patient must make to the center doing the clinical trials. Other investigators have shown that telemedicine can greatly improve side effect management in patients on chemotherapy.
Finally, nearly all patients have cell phones that contain a variety of sensors that are increasingly being used to monitor patient physiologic function. However, wearables like the Apple watch may have more promise then cell phones. Already, wearables have seen successful use in monitoring patients for cardiovascular disease and Parkinson’s disease.
This is an exciting time in the use of technology to improve patient care. However, we are only at the beginning of this revolution.