In July 2021, the creation of OracleBio’s dedicated Research & Development group marked an important milestone in our growth strategy.
In this article, we go behind the scenes and take a retrospective look at what we have achieved over the past year, and give you an insight into what the future holds.
The creation of the R&D group was motivated by 2 key drivers: 1) the willingness to stay at the forefront of technological innovations in the fields of digital pathology and image analysis, and 2) the desire to expand the portfolio of applications that we can offer to our clients.
The Digital Pathology arena is an extremely fast-paced environment, with constant innovations across the whole ecosystem, from advanced multiplex antibody panels and scanning technologies to cutting-edge Deep Learning architectures and paradigms.
Therefore, we found it imperative to create a dedicated team to monitor, assess, and ultimately test these novel tools and methods, where relevant to our business strategy.
Moreover, we also recognised the importance of not just being able to appraise new solutions but to have the technical capabilities to build bespoke applications too.
And that is how OracleBio’s R&D group was born — a multidisciplinary team with skills in software engineering, image and data analysis, histology, and biology.
The scientific scope of the R&D group includes the following areas:
- Technology and Platforms: evaluation of different imaging technologies (e.g. assays, scanners) and the different platforms that enable access to the generated images, such as image analysis software, database management, image ingestion processes or remote collaboration tools.
- Applications and programming: in-house software development capabilities, using programming languages such as Python or Matlab, to create new solutions to specific issues and/or adapt existing open-source programs to better fit the company’s needs.
- Consortia and External Collaborators: become stakeholders in UK/EU/worldwide digital pathology work groups, like the UK’s National Pathology Imaging Co-operative (NPIC) or the US-based Digital Pathology Association (DPA). Foster new collaborations with academic/clinical Centres of Excellence in Digital Pathology and maintain existing relationships with Universities. One prime example of a collaborative project we are involved in is the INCISE project (see our previous blog post here).
- Legislation: provide scientific advice and support to the company around regulatory frameworks (Good Clinical Practice, EU’s Artificial Intelligence Act…) and accredited standards, such as ISO 12052 Health Informatics – DICOM or ISO 13485 for Quality Management Systems.
The team’s project portfolio is a dynamic entity that grows and evolves according to the main priorities and objectives of the company. Although thematically diverse, projects can be divided into two main categories.
Reactive projects are those that emerge from an immediate need or issue, are relatively specific to a client or to a particular image analysis project, and can generally be completed in 1 to 3 weeks.
Some examples of reactive projects would be:
- Conversion of MS Excel spreadsheets containing analysis readouts into a client-specific file format, to make it compatible with their database management system.
- Conversion of image annotation overlays from a proprietary system to an open-source format, to enable a seamless switch between different image analysis software platforms.
- Creation of an app for generating a unique hash value (checksum) for each file present in a directory, to ensure data integrity during transfer.
Proactive projects, on the other hand, are long-term projects that have a global impact on most image analysis processes and align with OracleBio’s strategic objectives.
We are currently working on two key areas:
1) Automatic Image Quality Control (QC)
The first step image analysis scientists always need to perform before starting a study is to check that the quality of the images is good enough for analysis. That is, the image is in focus, free of artefacts, the staining is adequate and uniform, et cetera (check out our article series on image QC here).
However, doing these checks manually is a tedious and time-consuming job, especially in multiplex images with several channels. To streamline this process, we are currently building an AI-powered tool, developed in Python, that will automatically screen incoming images and highlight areas that are out of focus or that present tissue folds so that our scientists can automatically discard them and focus the analysis on viable tissue regions.
We are also testing the feasibility of a tool for assessing intra-batch stain variability, which will automatically flag over- and under-stained sections.
2) Spatial Analysis
Spatial omics is undoubtedly a powerful tool for better understanding the mechanisms of certain diseases, such as cancer, and how specific treatments interact with the tumour microenvironment.
As a consequence, more and more clients are asking for spatial readouts from their histology sections.
To cope with demand, last year we created our own ‘PhenoXplore’ tool, a Python app that enables us to identify any specific cellular phenotypes from object data generated with our image analysis software, and calculate a variety of spatial measurements on these, such as mean distance between populations, or a number of objects within a specified radius of another object.
We are currently working on expanding the catalogue of spatial measurements on offer.
Due to their length and impact, proactive projects need to be properly planned and managed throughout their lifecycle.
The diagram below illustrates this workflow:
The R&D project lifecycle’s 7 stages are:
- 1. OracleBio Senior Management Strategy and Vision: proactive projects stem from unmet needs identified by the company’s Senior Management Team, and are relevant to our mission, vision, and strategy.
- 2. Project Planning and Lifecycle Management plan: once a project has been identified, a project plan is drafted, which includes the scope of the work, feasibility, prioritisation, stakeholders, key deliverables, and timelines. The strategic impact of the project is also assessed, which is reflected in the lifecycle management plan.
- 3. Platforms and Architecture: the appropriate software is chosen depending on the specific objectives of the project. These platforms can range from off-the-shelf image analysis software to bespoke scripting in a variety of programming languages, such as Python, Matlab, or R script.
- 4. Development: the project functionality is built at this stage (which is usually the most fun part too!).
- 5. Testing and Validation: all deliverables produced are tested against a battery of mock datasets, to ensure the outputs generated by the software are consistent with the results/behaviour we expect.
- 6. Rollout, Training, and Support: once validated, the app is deployed into our production environment, and staff are trained on how to use it.
- 7. Lifecycle Management & Feedback: as part of our Continuous Improvement approach, we regularly review active applications and collect feedback from all project stakeholders. This information is passed on to Senior Management, where the project lifecycle workflow starts again.
It has been an exciting year for the R&D team, growing our internal capabilities, completing and delivering our first projects, and working with OracleBio’s Senior Management to devise a solid vision and strategy for the group.
The main focus for the next 3 months will be to test the robustness of the out-of-focus and fold detection prototypes, to ensure generalisation on images from different modalities and stains.
Next, the modules will be integrated into our AWS image ingestion system, such that these image QC routines will be automatically invoked every time we receive images from a client.
Also, from a process improvement perspective, we will be implementing a ‘DevOps’ framework for managing the lifecycle of software projects. We have recently implemented a version control system to track any changes that we make to our in-house developed programming scripts.
This is the first step toward developing a software lifecycle management framework, but there is still a long way to go.
The goal for year 2 will be to introduce functionality and integration testing, to automatically check that any changes we make to our apps in the production pipeline don’t adversely affect the overall behaviour of the system.
The integration of an R&D group, working alongside our image analysis scientists, will enable OracleBio to continue to deliver high-quality services and ensure we keep pace with the dynamic and exciting developments within digital pathology.
About the author:
Gabriel Reines March — R&D Project Manager at OracleBio
With a PhD in Biomedical Image Processing and a background in Electrical Engineering, Gabriel is a key member of the OracleBio R&D team. He oversees and manages the group’s project pipeline, and ensures that the company stays at the bleeding edge of the industry.
Find out more about OracleBio's Pathology Expertise
OracleBio's Quantitative Digital Pathology Services
Stay up to date with OracleBio
Sign up to our mailing list to stay informed about news, webinars, events and more