Open Research Questions — Unsolved Challenges

Science Advances by Naming What It Does Not Know

CROWN’s innovation programme is built on honest assessment of what has been achieved and what remains unsolved. The following research questions represent genuine engineering and methodological challenges that CROWN is actively working on — in dialogue with ETH Zürich and the broader research community.

Each question is a potential thesis topic, a possible grant application, or a starting point for academic collaboration. We publish them here because we believe that transparency about unsolved problems attracts better solutions than concealing them.

How can optical micro-imaging, near-infrared spectroscopy, and impedance sensing be integrated into a miniaturised tabletop platform without compromising measurement precision?

The CROWN Diagnostic currently requires four distinct sensor subsystems, each with its own optical path, signal conditioning electronics, and calibration requirements. Integrating these into a single compact platform presents challenges in thermal management (NIR sources generate heat that affects impedance measurements), optical isolation (imaging and spectroscopic light paths must not interfere), and mechanical stability (vibration tolerance for micro-imaging at 200x to 400x magnification).

Research in this area would draw on microsystems engineering, optical design, and embedded systems development. A successful outcome would be a reference design for a tabletop-scale platform that maintains measurement specifications equivalent to the current laboratory prototype.

Cross-Ethnic Classification Accuracy

How do we build an AI classification model that is equally accurate across all ethnic hair types when initial training data is necessarily limited?

The AI classification engine must achieve equivalent performance for the full diversity of human hair — from the straightest Type 1A through the tightest Type 4C coil, and everything in between and beyond those categories. Early-stage models will be trained on datasets that are smaller and less diverse than the eventual CROWN Hair Commons.

Research in this area would explore transfer learning from related domains (textile fibre analysis, material spectroscopy), synthetic data augmentation strategies, active learning for targeted data collection, and fairness-constrained optimisation that treats cross-category accuracy parity as a hard constraint rather than a soft objective.

Non-Destructive Chemical Treatment Detection

How can near-infrared spectroscopy reliably detect and characterise chemical treatment history in human hair without sample preparation or destruction?

Chemical treatment detection is one of the most socially significant measurements the CROWN Diagnostic performs. Evidence of chemical straightening, relaxing, or perming is direct, measurable data on conformity pressure — information that connects individual hair properties to population-level discrimination patterns.

The challenge is analytical chemistry at the practical level. Chemical relaxers (typically sodium hydroxide or guanidine hydroxide) permanently disrupt disulfide bonds in keratin, producing spectral changes in the 1100 to 1300 nm region. However, these changes must be distinguished from natural variation in protein structure, environmental damage, and the effects of other cosmetic treatments (conditioning, colouring, heat styling). The spectral signatures overlap, and the magnitude of change varies with treatment intensity and time elapsed since application.

Research in this area would involve controlled studies with known treatment histories, chemometric method development, and validation against established laboratory techniques (amino acid analysis, differential scanning calorimetry).

Privacy-Preserving Data Aggregation

How do we design data aggregation protocols for sensitive biometric hair data that enable population-level research while protecting individual privacy?

Every CROWN Hair DNA profile contributed to the CROWN Hair Commons contains biometric information that, in combination with demographic metadata, could potentially identify individuals. CROWN’s commitment to GDPR and Swiss nDSG compliance requires privacy protections that go beyond simple anonymisation.

Research in this area would explore differential privacy mechanisms (adding calibrated noise to aggregated statistics), federated learning approaches (training AI models without centralising raw data), secure multi-party computation (enabling cross-institutional analysis without data sharing), and k-anonymity or l-diversity guarantees for published datasets. The challenge is to implement these protections without degrading the statistical utility of the data for research purposes.

Sensor Calibration Standardisation

How can sensor calibration be standardised across CROWN Diagnostic devices operating in different environments, ensuring that measurements are comparable across sites?

The CROWN Hair Commons aggregates data from multiple devices in multiple locations. For this data to support valid cross-site comparisons, each device must produce equivalent measurements for the same sample. This requires calibration protocols that account for environmental variation (temperature, humidity, ambient light), component ageing (LED intensity decay, sensor drift), and operator differences (sample positioning, measurement duration).

Research in this area would draw on metrology — the science of measurement — and would involve developing reference standards (synthetic hair samples with known, stable properties), calibration verification procedures, and statistical methods for detecting and correcting inter-device bias in aggregated datasets.

Longitudinal Measurement Consistency

How do we track meaningful changes in an individual’s hair properties over time while distinguishing genuine change from measurement noise?

CROWN Hair DNA profiles have the potential to track changes in hair condition longitudinally — monitoring the effects of treatment changes, environmental exposure, or the cessation of chemical straightening. However, hair properties vary naturally across the scalp, across seasons, and across washing cycles. Distinguishing a genuine trend from natural fluctuation requires repeated measurements, statistical modelling, and careful experimental design.

Research in this area would involve designing measurement protocols for longitudinal studies, developing statistical models for within-individual variation, and validating the sensitivity of CROWN’s sensor modalities to clinically meaningful changes in hair condition.

An Invitation

These questions are not exhaustive. New challenges emerge as the research progresses, and solutions to one problem frequently reveal the next. CROWN publishes this catalogue as an invitation to the research community: if any of these questions interests you, we would welcome a conversation about collaboration.

For ETH Zürich students, several of these questions are suitable for semester projects or master’s theses. For established researchers, they represent potential joint grant applications or co-supervised doctoral projects. For engineers in industry, they describe the technical frontier where CROWN is working.