Imaging is key to success in oncology translational research and drug development

By Peter A. Duncan, Director - Marketing & Business Development, Life Sciences, Definiens AG

As defined by the NCI, "Translational research transforms scientific discoveries arising from laboratory, clinical, or population studies into clinical applications to reduce cancer incidence, morbidity, and mortality." Biomarkers discovered and validated in oncology translational research are increasingly being utilized for disease prognostication, clinical trial enrichment, companion assays for targeted therapies, and surrogates for toxicity. Bio-pharmaceutical companies are prioritizing their drug development efforts through the use of biomarkers that have been identified and validated through translational research, and this trend is likely to increase.

In order to be successful in both translational research and drug development, scientists need to compare the pathology of diseases in animals with that of humans; and to measure the efficacy of treatment, or the effects of toxicity. Molecular imaging and cell-based assays give early indications in pre-clinical research of the effectiveness or toxicity of therapeutics. Images acquired from tissue samples and non-invasive scans (CT, MRI, etc.) reveal valuable insights about a disease, its progression, and whether or not a patient is responding.

As a consequence, bio-pharmaceutical companies developing new drugs need to capture and assemble not only molecular and genomic information, but also complex, heterogeneous image data from all stages of drug development. As a result, image analysis becomes vital in the drug development process.

The image analysis challenge
As much as 70 per cent of all data generated in the life sciences is in image form and the volume is increasing. High-throughput image acquisition devices in the laboratory can produce thousands of images a day. Today, the task of analysis falls largely to experts in a specific therapeutic area who draw on their experience. This manual process is slow and, to a considerable degree, subjective.

There is a pressing need for automated image analysis solutions that work across the biomedical continuum to bridge the gap between the laboratory work bench and the patient’s bedside. However, despite decades of research and development, automated image analysis solutions have fallen far short of their promise. Additionally, the proliferation of imaging systems and proprietary image analysis tools creates vast complexity, which is difficult and costly to manage. Proprietary image analysis tools can often only be modified by their vendors, which delays research and increases costs. A platform independent, enterprise deployable, integrated image intelligence system is needed to make the most out of precious image data.

Definiens Cognition Network Technology
Definiens offers the world’s most advanced technology for extracting intelligence from images. The underlying technology was developed by Dr. Gerd Binnig, the 1986 Nobel Laureate for Physics, and his team to emulate human cognitive processes. The image analysis system looks at pictures in the same way as humans do – this is called Definiens Cognition Network Technology.

Instead of examining an image pixel by pixel or using pattern recognition methodology, Definiens’ segmentation and classification processes logically assemble groups of pixels as objects, according to characteristic shapes, colors and textures. Definiens Cognition Network Technology examines objects in relationship to each other and understands scale, overlapping objects and the relationship of two-dimensional images to three-dimensional shapes. In this way, Definiens software not only recognizes patterns, but more importantly, succeeds when there seems to be no pattern whatsoever; which is where critical biological information can often be found to facilitate better insight.

The data that results is multi-parametric, and relational, providing the scientist with a comprehensive, quantitative understanding of what originally presented itself as a complex image. The technology can be used with cell-based assays, histology specimens and medical imaging scans; in 2D, 3D and 4D. Definiens’ technology delivers far more accurate results than manual image analysis, in a fraction of the time. Thousands of images can be analyzed with precision, easing the image analysis bottleneck. Additionally, Definiens provides a software platform which can analyze virtually any image generated from any source.

Definiens in histopathology
Definiens technology excels in histopathology applications, where it can speed up analysis, reduce human error and help lab specialists make better decisions. For example, the technology can reliably indentify cancerous versus normal gland units and then quantify biomarkers within the context of microanatomy, including sub-cellular compartments. It measures a wide variety of morphological parameters, including spectral statistics, shape, size, position, texture and relationships to neighboring objects. Definiens TissueMap™ provides a solution for automatic nuclear, membrane, and cytoplasmic biomarker quantitation in whole tissue sections and tissue microarrays. Following are three case studies highlighting the utility of Definiens image analysis for histopathology applications:

Case Study: Her2/neu quantification for more accurate patient stratification
The accurate risk stratification of breast cancer patients using traditional immunohistochemical methods is a significant challenge, as patients can often be misclassified when antigen expression is borderline. Definiens has developed an approach to quantify membrane biomarkers such as Her2/neu whereby antigen expression is evaluated on a cell-by-cell basis. The software identifies both cell and membrane, before measuring the portion of stained area of the membrane; assigning 1+, 2+, or 3+ to each individual cell. This allows for far more precise assessment of antigen expression as compared to subjective, manual qualitative assessments, or other methodologies that rely on digital masking techniques.

In a recent study of 1800 patients from a multicenter breast cancer clinical trial for a leading targeted therapy, Definiens was able to demonstrate a 30% improvement in accuracy of assessing the expression of the Her2/neu antigen in breast cancer biopsies as compared to manual, qualitative pathology scoring methods. The impact of being able to more effectively enrich clinical trials by using a more accurate approach to assess expression levels of targets (or related antigens) would be felt, as less than optimal trial designs and / or enrollment often affect trial outcome in terms of whether or not primary or secondary endpoints are reached.

Case Study: Pre-clinical toxicity study accelerates FDA approval
A Definiens pharmaceutical partner faced an enormous challenge when the FDA required them to quantify the nuclear proliferation index within the jejunum tissue of rat animal models in which the drug was previously administered. Traditionally, this task had been performed by pathologists manually circling regions of interest with a mouse, and then manually or automatically counting nuclei that are stained for proliferation.

Definiens provided a solution that allowed repeatable and standardized counting of cell nuclei and quantification of crypt area – a daunting task because of the high heterogeneity of the tissue due to various crypt sizes and shapes. The toxicity study consisted of 180 rats and some 9000 images which were evaluated using Definiens automated image analysis and was completed in 4 weeks compared to the 16 weeks it would have taken with manual counting; saving time, millions of dollars in overhead costs, and thus facilitating a quicker approval by the FDA.

Case Study: Prostate cancer prognosis
A Definiens clinical services partner has developed a commercially available multivariate test which integrates clinical, molecular, and histological features from needle biopsies taken from the prostate cancer patient at the time of diagnosis. This information is used to generate a recurrence score for patients who will undergo radical prostatectomy. The test makes use of the quantification of histological features derived using Definiens software (for example, area of lumens) and integrates this information with localized biomarker quantification features also derived using Definiens software (for example, total area of androgen receptor expressed in alpha-Methylacyl-CoA racemase positive epithelial gland units). This approach provides the urologist a more accurate way to assess future risk of recurrence than existing methods of risk assessment, and could be an important factor in determining more effective treatment regimens.

Conclusion
Images are increasingly essential to translational research and oncology drug development. Definiens provides a world leading image analysis technology, built on an open platform that is massively scalable. It meets the image analysis needs of both researchers and clinicians, allowing them to work more efficiently, enabling a more complete understanding of clinically relevant biological questions. It provides a platform that spans all stages of drug development, supporting translational research and the drive towards personalized medicine.