Advanced Image Analysis Combats Neurodegenerative Disease

By Lina Sorg

The effective management of big data is becoming increasingly important within the applied math and computational science community, and analysis of datasets often yields patterns, industry trends, and relationships that reveal information about human behavior. The ability to quickly gather mass quantities of data has grown immensely in recent years, and successful analysis techniques are necessary to process this data and apply it to real-world situations. One such application is in the treatment of diseases and disorders, such as neurodegenerative disease. 

In a minisymposium presentation titled “Big Data in High Throughput Screening” at the SIAM Annual Meeting, Steven Finkbeiner (University of California, San Francisco) explains his use of image analysis and advanced technology to better understand neurodegenerative diseases such as Parkinson's, Alzheimer's, and Huntington's. “Neurodegenerative disease really stands out as one where we’ve made no progress,” he said, “and the instances of these cases are only getting longer.” Finkbeiner added that as of 2016, there are still no strong therapies that truly slow the disease, though not for lack of trying; numerous treatments have looked promising in mouse cells, but failed to deliver in humans. “I think as a community we’ve gotten really good at making safe but ineffective drugs,” he said. “This is devastating for patients and families, but also for drug companies who spend much money on late-stage testing.”

To combat these setbacks, scientists need to develop a more predictive preclinical pipeline that determines what compounds will generate human results. “Neurodegeneration itself is a challenging thing to study,” Finkbeiner admitted, because of its stochastic quality; it affects some cells more acutely than others and acts on varied time scales. 

Innovative scientific imaging technology enables researchers to create detailed biological pictures of cells, and Finkbeiner’s research group is utilizing machine learning technology to better understand neurodegeneration. With the help of a technology company, they have created specialized robotic microscopes that amass and examine biological images from stem cells more quickly and effectively than ever before. 

The fully-automated microscopes collect single-cell imaging data at a constant rate that allows for both phenotypic screening and the screening of small molecules. The microscopes produce ‘movies’ of live neurons (identified with fluorescent dye) that follow the growth and decay of cells in real time, measure and compare changes in neurons, and introduce different sensors to obtain varied data. The cells are also sparsely labeled to facilitate automated image analysis. “We can measure changes that occur during that cell’s lifetime,” Finkbeiner said. “We’re using certain statistical techniques to determine what factors are unfolding during a cell’s lifetime that will predict what will happen in the future.”

Most screening technologies only capture images at a single point, but Finkbeiner’s microscopes are between 100 and 1,000 times more sensitive. They also generate over 10 terabytes of data—hundreds of thousands of images—daily. As a result, the group’s imaging efforts have attracted the attention of multiple pharmaceutical companies. “More people are interested than we have hands to do it,” he said. “And we’re generating data faster than we can analyze it. That’s a fundamental problem for us right now.”

Ultimately, Finkbeiner hopes to achieve an explicit understanding of the underlying biology that drives neural activity. Tracking neurodegenerative cells and neurons and identifying harmful, inconsequential, and beneficial changes puts researchers one step closer to the creation of a blueprint explaining neurodegenerative disease. 

Lina Sorg is the associate editor of SIAM News.