Detection of the Onset of Multiple Sclerosis from SD-OCT Thickness Measurements of the Retinal Nerve Fiber Layer

ABSTRACT

The cumulative sum (CUSUM) likelihood ratio detector (discussed in Lai [1995] and Frisen [2003]) can be used to monitor a process level against linear trend changes. In this article we discuss this detector and its relationship to the standard CUSUM, which guards against step changes in the level; provide software for its implementation; and present a case study that uses this method to detect the onset of multiple sclerosis.

The thickness of the retinal nerve fiber layer has been suggested as a diagnostic measure for evaluating the progression of multiple sclerosis. Newly improved optical coherence tomography (OCT) instrumentation using scan averaging and alignment with a reference scan has reduced the short-term measurement variability in the retinal nerve fiber layer thickness to an extent that allows the proposed CUSUM likelihood ratio detector to detect clinically relevant changes over time. We show that this procedure detects such changes quickly and performs better than the traditional CUSUM technique that protects against step changes in the level.

KEYWORDS:

• cumulative sum likelihood ratio detector
• CUSUM
• statistical process control
• trend changes

ACKNOWLEDGMENT

This research was supported by a VA Rehabilitation Research and Development Center of Excellence Grant. We acknowledge the helpful comments of a referee that helped strengthen the focus of the article.

Additional information

Notes on contributors

Johannes Ledolter

Dr. Johannes Ledolter is the Maxwell Stanley Professor of International Operations Management at the University of Iowa. He is a fellow of ASQ and ASA. His email address is [email protected].

Randy Kardon

Dr. Randy Kardon is the Pomerantz Family Chair Professor in Ophthalmology and the Director of the Veterans Affairs Medical Center for the Prevention and Treatment of Visual Loss at the University of Iowa. His email address is [email protected].