Artificial Intelligence (AI) now permeates all aspects of daily life, including the field of medicine. AI and its specific components, Machine Learning and Deep Learning, are silent assistants in healthcare administration, research, and clinical practices from diagnostics to decision-making. This webpage showcases current applications of medical AI and its potential for transforming medicine, as well as Dr. Sperling’s involvement in utilization.
What is Artificial Intelligence (AI)?
AI draws upon the neural map of the human brain. Machines—i.e., computers and software—are modeled accordingly, with programs trained to identify, understand, analyze, self-teach and make predictions about a given dataset based on statistical algorithms.
Machines are trained on input from large datasets that are structured into machine language, so that computers operate as our brains do from infancy. They learn to recognize content such as patterns and features. For example, AI can be trained to distinguish characteristics of cancer cells that may be invisible to the naked eye, and make judgements about millions of bits of new data much faster than experienced pathologists working with powerful microscopes. This can aid human efficiency and reduce the workload of reading and interpreting slides, though humans still make final decisions.
Deep learning programs are designed to generate ongoing autolearning without human input. These man-made neural networks are modeled on how the human brain learns through experience. Deep learning anticipates new information, and makes accurate sense of incoming but previously unfamiliar input.
AI at work in clinical practice
Let’s start by briefly acknowledging the many AI resources that already streamline the business administration of medical practice. A noteworthy example is electronic health records (EHR): millions of case files are stored in the cloud and shared among medical centers and caregiver teams while protecting patient information as mandated by HIPAA.
Administration aside, this site is devoted to key examples of AI in clinical practice.
- Computer-aided detection (CAD). Early detection of cancer saves lives. Studies show that deep learning-driven CAD can achieve “…a cancer detection accuracy comparable to an average breast radiologist…”[i] A recent comparison of three AI algorithms used to identify breast cancer from screening mammograms found that one radiologist plus the best-performing CAD were together more accurate than two radiologists combined using no CAD.
- Enhanced diagnosis. Radiomics is a new field that merges medical scans (MRI, CT, PET/CT) and mathematical analysis. Images’ textural factors, e.g., signal intensity, pixels/voxels, etc. form quantitative maps of multiple tissue properties that are analyzed by machine learning. The results can enhance interpretation, diagnosis and decision-making.
- Predicting catastrophic medical events. Deep Learning can extract individual patient information such as fat location and composition to foresee the likelihood of life-threatening events like heart attack or stroke. Such knowledge can help patients commit to prevention through treatment and constructive lifestyle changes. Another example is foreseeing and diagnostically validating escalation in mental illness based on schizophrenic patients’ communication style in social media posts.
- Liberating more time with patients. More accurate and efficient diagnostic information can expand and transcend current clinical methods. Access to EHR makes shared patient records more efficient across multiple systems. Thus, AI has the potential to liberate a doctor’s time in order to be more personally available and energized for each patient; it can also be harnessed to improve a doctor’s ability to communicate with patients by providing teaching and self-evaluative “bedside manner” tools.[ii]
The importance of experience
Considering ever-developing advances in computational power and availability of large data sets, the marriage of humans and machines in future clinical practice seems inevitable. Therefore, regardless of their feelings, the radiologist should be familiar with these concepts.[iii]
The above quote hints that radiologists with expertise in the evolution and contemporary utilization of AI in medicine will have a leading edge as the field is rapidly growing. AI is taking medicine from the ordinary to the extraordinary on so many levels.
Dr. Dan Sperling has been ahead of this curve for nearly a decade. He has immersed himself in radiologic applications of all aspects of AI. He and his staff have integrated AI tools into their image-based detection/diagnostic technologies, and applied it in post-treatment (Focal Laser Ablation) follow-up in order to maximize clinical benefits for prostate cancer patients (Sperling Prostate Center), neurology patients (Sperling Neurosurgery Associates), and all who receive imaging services (Sperling Medical Group).
His areas of expertise include:
- Novel quantitative image analysis (development, evaluation, application)
- Multi-modal co-registration tools
- Machine learning tools
- Practical application (e.g., targeted prostate biopsy, Focused Ultrasound thalamotomy)
Quantifying Post- Laser Ablation Prostate Therapy Changes on MRI via a Domain-Specific Biomechanical Model: Preliminary Findings. PLoS ONE. 2016 Apr; 11(4): e0150016. DOI: 10.1371/journal.pone.0150016
Identifying Quantitative In Vivo Multi-Parametric MRI Features for Treatment Related Changes after Laser Interstitial Thermal Therapy of Prostate Cancer. Neurocomputing. 2014 Nov 20; 144: 13-23. DOI: 10.1016/j.neucom.2014.03.065
Commercially Available 3D Co-Registration Systems. Presenter, Advances in Prostate Imaging and Focal Ablative Treatment of Prostate Cancer physician course, NYU School of Medicine/Langone Medical Center, New York NY, June 2014.
Quantitative Evaluation of Treatment Related Changes on Multi-Parametric MRI after Laser Interstitial Thermal Therapy of Prostate Cancer. Proc Soc Photo Opt Instrum Eng. 2013 Mar 15;8671:86711F.
Simultaneous Segmentation of Prostatic Zones Using Active Appearance Models With Multiple Coupled Levelsets. Computer Vision and Image Understanding. 2013 Aug; 117(9):1051-1060. DOI: 10.1016/j.cviu.2012.11.013
Prostate Cancer Imaging. Plenary speaker, workshop for physicians. Medical Image Computing and Computer Assisted Intervention, 14th Annual Conference, Toronto ONT. September 2011.
Academic Industrial Clinical Partnerships in Prostate Imaging and Image Guided Interventions. Panelist. Medical Image Computing and Computer Assisted Intervention, 14th Annual Conference, Toronto ONT. September 2011.
The Future of Prostate Cancer Detection and Treatment Education and Guidance for Radiology Technologists. Advanced Radiology. May 2011.
[i] Rodriguez-Ruiz A, Lång K, Gubern-Merida A, Broeders M et al. Stand-Alone Artificial Intelligence for Breast Cancer Detection in Mammography: Comparison With 101 Radiologists. JNCI. 2019 Sep; 111(9): 916-922.
[ii] Rowe, J. “What AI Could Do for Doctor-Patient Relationships.” AI Powered Healthcare, Aug. 29, 2019.
[iii] Koçak B, Durmaz E, Ates E, K?l?çkesmez Ö. Radiomics with artificial intelligence: a practical guide for beginners. Diagn Interv Radiol. 2019 Nov; 25(6): 485–495