Aegis Digital Technologies is developing Micro-SDP, a mathematically rigorous, deterministic AI platform for automated detection and classification of single blood cells in microscopic images, targeting cancer monitoring and clinical diagnostics.
Smart Detection & Prediction of Single Cells in Microscopic Images
Micro-SDP is an AI-based Software as a Medical Device (SaMD) for automated detection and classification of single blood cells in heterogeneous microscopic images, built on a hybrid architecture: learning algorithms and model updates are computed server-side on Aegis infrastructure, while all patient image analysis is performed locally on the clinical device. Its core mission is to enable continuous, data-driven monitoring of cancer-relevant cell populations, with the goal of giving clinicians a faster, reproducible, and auditable tool for oncology diagnostics.
Unlike probabilistic deep learning approaches, Micro-SDP is built on a deterministic variational mathematical framework, specifically, a Primal-Dual Hybrid Gradient (PDHG) optimisation algorithm applied to dictionary learning. This yields mathematically proven existence, uniqueness, and stability guarantees that are essential for clinical-grade reproducibility and regulatory traceability.
The following 10-minute overview of Micro-SDP explains what it does, how it works, and where it is headed.
One of the core learning algorithms of Micro-SDP is developed and tested on the Berkeley Single Cell Computational Microscopy (BSCCM) dataset [Pinkard et al., 2024], openly provided by the computational imaging laboratory of UC Berkeley. The dataset comprises five imaging channels: DPC Left, DPC Right, DPC Top, DPC Bottom, and Brightfield. The algorithm can be accessed as a preprint on Zenodo in the link below.
Variational Dictionary Learning with Hybrid ℓ1 and Non-Negativity Penalization for Single-Cell Microscopy
Erdem Altuntaç, PhD | Aegis Digital Technologies, Dresden
Micro-SDP is being developed in alignment with EU IVDR 2017/746 (In Vitro Diagnostic Regulation) and FDA Software as a Medical Device (SaMD) guidelines, targeting Class C IVD classification under the IVDR framework.
Aegis Digital Technologies is a deep-tech startup founded by Dr. Erdem Altuntaç in Dresden, Germany, in 2024. The company's sole focus is currently the development of Micro-SDP, an AI platform for automated single-cell detection and classification in microscopic blood images, targeting oncology and diagnostics.
Dr. Erdem Altuntaç holds a PhD in Applied Mathematics from the University of Göttingen (2016), where his research focused on iterative variational regularisation and optimisation algorithms for inverse ill-posed problems. His postdoctoral work at the Université Libre de Bruxelles (2017-2019) introduced him to primal-dual splitting methods, the mathematical foundation that directly underpins Micro-SDP's core algorithm. At Fraunhofer FHR (2019-2021), he applied these methods to radar signal processing and LiDAR-based depth completion in automotive perception systems. His subsequent postdoctoral position at DZNE Dresden (2021-2024) brought him into biomedical signal processing, where he developed predictive models for physiological time-series data using deep learning. This trajectory, from inverse problems and variational mathematics, through industrial signal processing, to biomedical AI, converges directly in Micro-SDP. The human motivation behind the project is deeply personal: some family experiences with misdiagnosis and its consequences drives the commitment to building diagnostic tools that are rigorous, reproducible, and trustworthy.
The name Aegis, the shield of protection in Greek mythology, reflects the founding principle of Aegis Digital Technologies: to build technology that protects. Whether in medical diagnostics, clinical decision support, or any future domain where software touches human outcomes, our commitment remains the same. Determinism, transparency, and reproducibility are not design preferences, they are non-negotiable requirements for systems where the cost of failure is measured in human lives.
For collaboration, investment, or clinical partnership enquiries:
Email: [email protected]
Dresden, Germany