Utility of 3D Facial Analysis As A Biomarker In Rare Diseases Exploration with Hereditary Angioedema

This multicentre observational study in Singapore demonstrates that 3D facial analysis powered by AI can identify distinct digital biomarkers, such as periorbital swelling, to monitor flare-ups in Hereditary Angioedema, offering a promising non-invasive tool for disease management in rare conditions.

The Gist

Imagine you have a very rare, tricky puzzle to solve. For people with Rare Diseases, finding the right piece of the puzzle (a diagnosis) can take years, often leaving them in a state of confusion and anxiety known as the "diagnostic odyssey."

This paper is like a story about a team of detectives trying to solve these puzzles faster using a new, high-tech magnifying glass: 3D Facial Photography.

Here is the breakdown of their adventure in simple terms:

1. The Problem: The "Needle in a Haystack"

There are over 10,000 rare diseases. It's impossible for any single doctor to know the face of every single one. Often, the symptoms of a rare disease look just like common ailments, leading to misdiagnoses.

• The Metaphor: Imagine trying to find a specific, unique snowflake in a blizzard. Most doctors see "snow," but they miss the unique pattern of the specific snowflake they are looking for.

2. The New Tool: The "Digital Mirror"

The researchers decided to use Digital Phenotyping. Instead of just looking at a patient with their eyes, they took 3D photos of their faces.

• The Analogy: Think of a standard photo as a flat drawing. A 3D photo is like a sculpture. It captures the exact height, depth, and curves of a face, down to the millimeter.
• They used a special camera (Vectra M3) and smart software (Cliniface) that acts like a super-precise ruler, measuring tiny details of the face that the human eye might miss.

3. The Target: Hereditary Angioedema (HAE)

The team focused on a specific rare disease called Hereditary Angioedema (HAE).

• What is it? It's a condition where people get sudden, unpredictable swelling (like a balloon inflating) in their face, throat, or stomach.
• The Challenge: The swelling comes and goes. When it's gone, the face looks normal. When it's there, it looks puffy. Doctors often struggle to know if a patient is having a "flare-up" or just a normal day, especially if the swelling is mild.
• The Goal: Could this 3D camera act as a digital biomarker? (A "biomarker" is just a fancy word for a clue that proves a disease is present). Could the camera see the swelling before the patient even feels it?

4. The Experiment: The "Face-Off"

The team gathered two groups of people in Singapore:

1. The Test Group: 20 people with rare diseases (including 7 with HAE).
2. The Control Group: A massive database of "normal" faces from people of similar age and ethnicity (Chinese, Malay, Indian) to serve as a baseline.

They took 3D scans of everyone and let the computer compare the "Test Group" faces against the "Normal" faces.

5. The Findings: What the Camera Saw

The results were a mix of "We found something!" and "We need more data."

• The "Smoking Gun": The computer found one specific measurement that stood out: the angle around the outer corner of the eye and the nose. In the rare disease group, this angle was statistically different from the normal group.
  • Simple translation: The "face map" of these patients had a unique signature that the computer could spot, even when the doctors couldn't see it with the naked eye.
• The "Time-Lapse" Magic: For two patients with HAE, the team took photos over a long period (months and years).
  • Patient 1: The camera caught a swelling of more than 3mm around the right eye right after a flare-up. It even tracked how the swelling went down when the patient took medicine, and how it came back when the patient forgot to take their pills!
  • Patient 2: The camera spotted that the swelling wasn't just "puffy"; it was asymmetric (one side was puffier than the other).
  • The Metaphor: It's like having a security camera that doesn't just record a crime, but measures exactly how much the burglar moved the furniture, even if the burglar tried to hide it.

6. Why This Matters

If this technology works perfectly, it could change the game for rare diseases:

• Faster Diagnosis: Instead of waiting years to figure out what's wrong, a 3D scan could give a doctor a "red flag" that says, "Hey, this face looks like it belongs to a specific rare disease."
• Better Monitoring: For HAE patients, they could scan their faces at home or in the clinic to see if a flare-up is starting before it becomes dangerous. It removes the guesswork.
• Objective Proof: Currently, doctors rely on patients saying, "I feel swollen." This gives them a number: "Your face is 3mm puffier than yesterday."

7. The Catch (Limitations)

The researchers are honest: This was a small study.

• They only had 20 people in the main group and just 2 people they watched closely over time.
• It's like testing a new car engine on a short track with only two drivers. It looks promising, but they need to test it on a highway with thousands of drivers to be sure it works for everyone.
• They need more data from different ethnic groups and ages to make the "rulebook" complete.

The Bottom Line

This paper is a proof of concept. It's the first step in saying, "Hey, our faces hold secret codes for our diseases, and 3D cameras + AI might be the key to reading them."

It's not a magic wand yet, but it's a very shiny, promising new tool that could one day help people with rare diseases skip the long, scary "diagnostic odyssey" and get the help they need much faster.

Technical Summary

1. Problem Statement

People living with rare diseases (PLWRD) frequently face a "diagnostic odyssey," often lasting over four years, due to the sheer number of rare conditions (>10,000) and overlapping phenotypic features. Specifically, Hereditary Angioedema (HAE) presents a unique challenge:

• Lack of Objective Biomarkers: HAE is characterized by intermittent, transient soft-tissue swelling (oedema), particularly in the periorbital and perioral regions. However, current diagnosis and monitoring rely heavily on subjective patient reporting and clinical observation.
• Delayed Treatment: Without objective markers to define exacerbations, patients often receive delayed treatment, leading to increased anxiety and life-threatening airway complications.
• Clinical Trial Limitations: Rare disease trials suffer from small sample sizes, high heterogeneity, and subjective outcome measures, leading to high failure rates. There is a critical need for non-invasive, objective Digital Biomarkers (DBMs) to monitor disease progression and treatment response.

2. Methodology

This was a multicentre, prospective observational study conducted across three hospitals in Singapore over 4 years and 8 months. The study utilized Digital Phenotyping (DP) via 3D facial photogrammetry.

Study Design & Cohorts

The study employed two complementary approaches:

1. Quantitative Case-Control Analysis:
   • Test Cohort: 20 participants of Chinese genetic ancestry (5 with HAE, 15 with other Rare Diseases exhibiting facial features like oedema or asymmetry, e.g., Mucopolysaccharidosis, Achondroplasia, Noonan syndrome).
   • Control Cohort: A reference dataset of ethnically matched (Chinese) healthy individuals used to generate growth curves.
2. Qualitative Longitudinal Assessment:
   • Participants: Two specific HAE patients (one adult male of Malay ancestry, one adult female of Indian ancestry).
   • Protocol: Serial 3D imaging captured during baseline (non-acute) and acute flare-up stages over periods ranging from 261 to 687 days.

Data Acquisition & Processing

• Imaging: 3D facial scans were captured using the Vectra M3 camera system.
• Analysis Software: Images were processed using Cliniface (AI-powered software) to extract 58 base facial measurements, generating 108 distinct measurement variables (including distances, angles, and asymmetry metrics).
• Statistical Approach:
  • Growth curves were regressed for the control cohort (ages 0–70) to establish population means and standard deviations.
  • Z-scores were calculated for test cohort participants against these age-matched curves.
  • Hypothesis Testing: Two-tailed t-tests were performed to compare test cohort Z-scores against the null hypothesis (mean = 0).
  • Correction: The Holm-Bonferroni method was applied to control the Family-Wise Error Rate (FWER) across the 108 variables.

3. Key Contributions

• Proof of Concept for DBMs: Demonstrated that 3D facial analysis can identify specific Digital Biomarkers (DBMs) for HAE, specifically focusing on periorbital swelling and facial asymmetry.
• Objective Monitoring: Established a framework for using non-invasive 3D imaging to objectively differentiate between "flare" (acute) and "non-flare" (remission) states, as well as to detect subtle subclinical exacerbations.
• Asymmetry Detection: Highlighted that relying solely on symmetric landmarks may miss critical disease indicators; the study emphasized the importance of asymmetry metrics (e.g., unilateral peri-orbital swelling) as a key differentiator in HAE.

4. Results

Quantitative Findings (Case-Control)

• Statistical Significance: Out of 108 variables, 33 failed normality tests. Of the remaining 75, 11 had raw p-values < 0.05.
• Primary Biomarker: After FWER correction, only one variable remained statistically significant: Outer Canthal, Nasal Angle (Raw p = 0.000, Adjusted p = 0.010).
  • The test cohort showed a mean effect size (Z-score) of 0.87, indicating a deviation from the normative growth curve.
• Secondary Indicators: Ten other variables (including Interpupillary Distance, Cutaneous Lower Lip Height, and various asymmetry measures) showed small p-values (0.005–0.034) before correction, suggesting potential utility for monitoring facial oedema resolution or mild exacerbations, though they require further validation.

Qualitative Findings (Longitudinal HAE Cases)

• Participant 1 (Male, Malay): Serial imaging over 687 days revealed asymmetric peri-orbital swelling (>3mm increase around the right eye) during a flare. The 3D analysis successfully tracked the resolution of swelling and detected a mild, subclinical exacerbation linked to medication non-compliance.
• Participant 2 (Female, Indian): Imaging over 261 days showed pronounced asymmetric swelling on the left side (peri-orbital, cheek, and lower jaw).
• Key Insight: The study confirmed that 3D imaging can detect subtle shape changes and asymmetry that are often missed by standard clinical observation or symmetric landmark analysis alone.

5. Significance and Limitations

Significance

• Precision Medicine: This study validates the utility of 3D facial analysis as a non-invasive tool for precision medicine in rare diseases. It offers a pathway to shorten the diagnostic odyssey and improve disease management through objective monitoring.
• Clinical Trial Utility: The ability to generate objective DBMs could revolutionize rare disease clinical trials by providing reliable, quantitative endpoints for treatment response, reducing reliance on subjective patient reporting.
• Scalability: The approach (3D photography + AI) is potentially scalable and replicable in real-world clinical settings.

Limitations

• Sample Size: The study was limited by a small cohort (n=20 for quantitative, n=2 for longitudinal), which restricts statistical power and generalizability.
• Ethnic Diversity: The quantitative analysis was restricted to participants of Chinese genetic ancestry. The results may not apply to other ethnic groups without further validation.
• Heterogeneity: The test cohort included various rare diseases, not just HAE, which introduces variability in facial phenotypes.
• Data Sharing: Participant-level data cannot be shared to protect anonymity, limiting independent verification.

Conclusion

The study provides preliminary but compelling evidence that 3D facial analysis can serve as a robust Digital Biomarker for Hereditary Angioedema. By quantifying facial asymmetry and swelling, this technology offers a promising avenue for early detection, monitoring of disease flares, and assessment of treatment compliance, ultimately aiming to improve health outcomes for patients with rare diseases. Future work requires larger, multi-ethnic cohorts to validate these findings.