Machine Learning Domain Generalizer

A strategic plan for improving model generalizability when faced with images from different scanners or labs.

In 3 weeks, you’ll have a plan for getting your project back on track.

Apply Now

Does your machine learning model perform well on your training data but fail on images from a different source?

You’ve put a great deal of effort into gathering, annotating, and cleaning data. You’ve explored different machine learning models and found the one that works best on your images. But then you test it on images from a different scanner or lab and performance drops.

You can’t get annotated images for training from every possible source that you need your model to perform on. It’s just not possible. You’re stuck with this domain shift.

Perhaps you’ve tried some techniques like image augmentation or stain normalization and only seen a small improvement. What else can you do?

Train machine learning models that generalize to different labs and scanners

Imagine that your models can handle the change in data distribution. You can rely on their robust predictions for inference images from a variety of labs and scanners. And you have the tools to tackle similar challenges with future ML projects.

The impact on your business of a successful ML implementation is immense. Get your project back on track.

Get a customized plan for improving model generalizability

Using my proven process, I'll assess your project and design a strategic plan to improve model performance on images from a different distribution than your training images.

Results:

  • The ability to train a machine learning model that better generalizes to images from other scanners and labs
  • Decreased likelihood of wasting time on unsuccessful approaches
  • Confidence that you're incorporating the latest and greatest tools and techniques
  • Save time by using existing codebases, when possible, instead of implementing from scratch
  • Increased clarity on the scope of work

Here’s how it works:

Our meetings will take place over Zoom over the course of 3 weeks. You’ll start by filling out a questionnaire with some background information about your project. I’ll dive deeper into this in our kickoff meeting, examining the desired outcomes for your project and what you’ve tried so far. I’ll also give you an overview of the strategies that we’ll discuss in more detail in our subsequent meetings.

In our second meeting, I’ll go over each strategy in more detail with customized advice for your project. I’ll review the plan in our final session and answer any outstanding questions.

A check-in call is also included in case you have any follow up questions as your team begins implementing the plan. This call is to be completed within six weeks of report delivery.

Deliverables:

You will receive a strategic plan detailing the techniques we discussed and outlining the best solutions for your project. The report will include:

  • A summary of techniques for improving model generalizability
  • Recommendations for your project
  • Links to relevant papers and codebases

100% money-back guarantee!

If you don’t feel the ML Domain Generalizer is the right fit for you, just let me know at the end of our kickoff meeting and I’ll refund your payment in full.

Don’t just take my word for it...

Pixel Scientia provides valuable insights for Gestalt's image analysis product development. Heather has deep understanding of digital pathology and machine learning and their application to whole slide images. Her in-depth review of the current state of the art research has enabled Gestalt to rapidly focus our machine learning efforts on approaches that are yielding value for our pathologists and their patients.

Brian Napora, Gestalt Diagnostics, VP of Sales & Product Management

Heather's knowledge of the current state of the art within the digital pathology field is second to none. Consultation on best practice approaches for deep learning model classification performance and insight into digital research perspectives were very beneficial. Discussions of prior art enabled our team to focus on novel research and refine our current AI methodologies for clinical research.

Jenny Fitzgerald, Deciphex, Director of Clinical Research & Operations

Still have questions?

What happens after I apply?

I’ll schedule a short call to confirm that this package is suitable for your project. Then I’ll send over a short contract. Once you submit your payment, you’ll receive an email with a link to a questionnaire. At the end of the questionnaire you’ll be able to schedule our kickoff call.

Will you sign an NDA before we start?

The contract will include confidentiality and non-disclosure provisions.

We already have a great ML team. Why do we need you?

Maybe you don't need me. If your team has spare time to dig through the machine learning research and identify the best strategies for accommodating a domain shift, you might not.

But if your team is spending most of their time training algorithms, that's where I come in. It is my job to stay at the forefront of the field, specifically at the intersection of machine learning, computer vision, and pathology. Through literature reviews and observations across multiple projects, I develop best practices for applying ML to pathology images and tactics for adapting new advances.

You and your team know your data and specific application better than I ever will. In developing a strategic plan for improving your model, I build upon your team’s expertise and combine it with my knowledge of the field to get you on the path to a successful ML solution.

Our project is different from other applications. Is this a problem?

No. All projects in this space are different. This is why I focus specifically on pathology so that I can dig into research that's relevant to the field. I will be sure to assess the closest related publications. There may be some differences between these and your application, but it is essential for identifying the best solutions for your project.

What’s next after the plan? Can you help us with implementation?

After completing this plan, we have a few options. Your team could go off and implement the algorithms outlined independently. However, my clients have found that their projects are completed more efficiently with continuous guidance. You can sign up for my Accelerator monthly service and receive advice as your team handles the implementation. For teams requiring additional support, I can assist with the implementation.

Still not sure if this is the right package for you?

Schedule a free Strategy Session, and I’ll help you determine if this is a good fit.

Who are you, anyway?

Pixel Scientia Labs is led by me, Heather Couture. I have 16 years of experience in machine learning, 10 of those with applications to pathology. While I have no medical training, I do have a PhD in Computer Science and have published in top-tier computer vision and medical imaging venues. I write regularly on LinkedIn, for my newsletter Pathology ML Insights, and for a variety of trade publications. You may have heard me on podcasts or at conferences.

I help pathology startups accelerate their machine learning projects. I make use of the latest ML research to amplify their results and support their in-house team for the long term. My mission is to fight cancer with AI – and I do that by strengthening the ML component of my clients’ most impactful projects.

Availability is limited

I only take on one new client a month with this ML Domain Generalizer offer. Scheduling is first come, first served. The sooner you apply, the sooner you will get your ML project back on track.