Optical data analysis looks to clinical applications

Christine Amwake

I am privileged to work with a highly interdisciplinary team where engineers work together with life scientists to solve interesting biological questions and pursue clinical applications of our work. The engineering experts are led by Tom Baer from the Stanford Photonics Research Center, and we provide optical and analytical techniques to collect and analyze data. The life science experts, who determine the pertinent biological questions, are led by Dr. Renee Reijo Pera from the Stanford School of Medicine.

Previous work under the direction of the Dr. Reijo Pera and Dr. Baer correlated dark-field time-lapse imaging with gene expression profiling to conclude that success of the embryo to develop into the blastocyst stage can be reliably predicted within two days of fertilization, when the embryo is made up of only four cells and before embryonic gene activation (EGA), by using time-lapse optical imaging and an algorithm to detect three key features in the development cycle.¹ The three parameters that are used to determine successful development into the blastocyst stage are (i) the duration of the first cytokinesis (the very brief last step in mitosis that physically separates the two daughter cells); (ii) time interval between the first and second mitosis events; and (iii) the time interval between the second and third mitosis events. Software developed by their team of researchers was used to detect the cells via computer vision and to follow the shape of the cells and evaluate the three parameters. The figure shows an example of how the computer vision method detects the cells and their shapes and follows them through this critical stage of development. The outcomes of the algorithm applied to time lapse imaging were correlated to gene expression data, and successful blastocyst formation was found to be predictable with 93% sensitivity and specificity.

The bottom row shows results of the tracking algorithm applied to the top row images of a single embryo. Images were taken at 5-minute intervals, and only a sampling of the images is shown here.¹

In an extension to this study, it was found that the same time lapse imaging parameters can be used within tighter tolerances to predict ploidy versus aneuploidy with 100% sensitivity and 66% specificity.² Aneuploidy is a condition found in 50-80% of cleavage stage embryos, where an abnormal number of chromosomes are present in the embryo which are either not compatible with live birth or can cause conditions such as down syndrome. Fragmentation of the embryo was shown to be correlated with time lapse imaging to predict types of aneuploidies. 

These studies proved promising for applications in successful in vitro fertilization (IVF), since typical IVF methods do not accurately predict healthy embryos and usually multiple embryos are transplanted with negative consequences such as multiple births, the need for fetal reduction, and miscarriage.¹ To this end the PIs and some co-authors of the paper listed in reference 1 have created the company Auxogyn, which brings this time-lapse imaging and computer vision automated analysis to the clinic in their Early Embryo Viability Assessment (EEVA) System.

The current work of our group applies similar imaging and analysis techniques to study other biological phenomena in culture. Optical imaging is an enabling technology for looking at biological samples because it is less invasive than many other imaging techniques. In many of our applications, we image the cells repeatedly over periods of several days, so the imaging system must be designed to support this effort. I look forward to updating you on more of our work soon. Thanks for reading!

REFERENCES
1. C. C. Wong et al., Nat. Biotechnol., 28, 1115-1121 (2010).
2. S. L. Chavez et al., Nat. Comm., 3, 1251 (2012) .

CHRISTINE AMWAKE is a PhD candidate in Electrical Engineering at Stanford University under the advisement of Prof. Olav Solgaard. Her current research interest is applying optical imaging techniques to biomedical applications.

Optical treatment planning for interstitial photodynamic therapy

Tim Baran

A number of research groups, including Thomas Foster’s lab at the University of Rochester, have become interested in using photodynamic therapy (PDT) to treat cancers that are located deep within the body.

PDT is an emerging cancer therapy that uses a combination of light-sensitive drugs, known as photosensitizers, and targeted illumination to create photochemical reactions that result in the destruction of cancer cells. Since the treated area is limited by the penetration depth of the treatment light, PDT has typically been used to treat superficial malignancies of the skin and other easily accessible regions.   

To make it work within the body, the photosensitizer is delivered systemically or locally, and allowed to accumulate in the tumor. Under image guidance, cylindrical diffusing fibers are then inserted into the tumor to deliver the treatment light. This treatment light is usually delivered by a laser with wavelengths varying from 630-700 nm (and beyond), depending on the photosensitizer used. As the diffusers used can be as long as 5 cm for large tumors, the required laser power at the source can approach or exceed 1 W.

As interstitial PDT is often performed in regions where there is sensitive healthy tissue nearby, there is a need for careful treatment planning. Towards that end, we have developed a Monte Carlo simulation space that allows for patient optical properties and anatomy to be incorporated into a rigorous treatment plan. 

Unlike radiation therapy, in which radiation doses can be directly computed from CT scan data, calculations of optical dose require knowledge of the patient’s optical properties, which can vary among patients and even within a single patient. Therefore, spectroscopic determination of optical properties is required before a treatment plan can be formulated.

Simulated patient data showing the insertion of four cylindrical diffusing fibers for photodynamic therapy of head and neck cancers

A number of techniques exist to do this. In our case, we use a custom optical probe that is inserted into the treatment region and a Monte Carlo based fitting algorithm in order to extract optical properties. These extracted optical properties are then combined with CT image data from the patient in order to build an optical and anatomical map of the patient in our simulation space. The number of diffusers, and the amount of light delivered by them, can then be optimized using a constrained nonlinear optimization algorithm. This ensures that tumor tissues receive a physician-prescribed light dose, while damage to healthy tissue is minimized.

My work thus far has been preclinical, using simulated data sets and animal models. We are actively seeking to translate interstitial PDT into the clinic at the University of Rochester Medical Center for treatment of cholangiocarcinoma, cancers of the head and neck, and deep-tissue microbial infections.

TIM BARAN is a PhD candidate in the Institute of Optics at the University of Rochester (Rochester, NY).  His research in the Foster lab is related to optical dosimetry and treatment planning for interstitial photodynamic therapy, with an emphasis on the simulation of light propagation in tissue.

Students! Professors! Want your work represented in BioOptics World StudentView? Contact us for guidelines: leem@pennwell.com, barbarag@pennwell.com.