Master Thesis: Imaging analysis algorithms

Aplex Bio is at the forefront of advanced molecular testing, delivering platform technologies that expand capabilities in precision genomics, molecular diagnostics, and high-frequency biological screening. Hyperplex PCR™, the company’s next-generation platform, introduces a new paradigm in molecular measurement by combining multiplexed detection, molecular counting, and single-nucleotide discrimination in a single workflow. The platform delivers digital-like single-copy sensitivity while scaling to 100+ targets per well, addressing limitations in plex, precision, and cost-efficiency that constrain both traditional PCR and sequencing technologies. Aplex Bio is dedicated to advancing the next generation of molecular diagnostics and enabling technologies for population-scale applications that can contribute to improved healthcare outcomes worldwide. The company is headquartered in Stockholm, Sweden.

Background

During our hpPCR workflow, microscopic spots are imaged across multiple channels and then quantified and classified downstream. A single experiment can produce large amounts of image data. This creates a strong need for analysis methods that are computationally efficient.

The broad goal of this thesis is to improve our spot detection and quantification in microscopy images, with attention to accuracy and runtime performance.

Possible Projects

The exact project can be shaped together with the student, but directions we propose include the following:

Fast Detection Model
Develop and evaluate lightweight computational methods for single-class object detection and segmentation in large microscopy images.

This direction may involve classical image processing, machine learning, model optimization, and performance-aware scientific programming, with a focus on methods that are practical for large microscopy datasets.

Morphology-Aware Spot Analysis
Extract and analyze shape-related information from fluorescent spots in microscopy images to allow deeper postprocessing.

This direction connects naturally to both PSF-aware modeling and fast spot detection. It can be done through either a shape fitting approach or through Bayesian Inference using MCMC-methods on prior information.

Spot Deconvolution and PSF Fitting
Conduct work on methods for modeling the point spread function (PSF), and how that can be leveraged for detection and segmentation tasks.

This direction may involve optical physics, image processing, inverse problems, and numerical optimization. The student is NOT expected to know these topics beforehand.

A suitable candidate has a background in engineering physics, computer science, applied mathematics, electrical engineering, machine learning, or a similar technical field.

Depending on scope, the project will likely be mathematically and computationally heavy, so we are interested in someone who is comfortable with quantitative reasoning, scientific programming, and open-ended technical problems.

Useful skills include:

• Python and general programming experience
• Image processing
• Experience in applied mathematics or signal processing
• Machine learning theory and applications
• Numerical optimization
• Interest in performance-aware programming for large datasets

Experience with modern AI-assisted coding tools such as Claude Code, Codex, Cursor, etc, is considered a big plus. The goal is to use these tools to speed up prototyping, so that more time can be spent on modeling, testing, and evaluating ideas. And the student is highly encouraged to leverage such tools in this work.

What we offer

• An ambitious, innovative, inclusive and professional environment.
• Being part of building a cross-disciplinary technology at the frontier of its space.

Start Time: August/September

Applications:

Please apply here with your CV and a motivation letter on why you are interested in writing your exam thesis at us.