Hyperspectral Retinal Observations for the Cross-sectional Detection of Alzheimer's Disease
- Conditions
- Cognitive DeclineCognitive ImpairmentAlzheimer Disease, Early Onset
- Interventions
- Procedure: non-invasive hyperspectral retinoscopyProcedure: blood sampleDiagnostic Test: Test of cognitive ability on tablet computer with CoGNIT software
- Registration Number
- NCT05604183
- Lead Sponsor
- Mantis Photonics AB
- Brief Summary
Two devices will be tested in this research:
1. Mantis Photonics' hyperspectral camera for non-invasive retinal examination (i.e., a hardware medical device under investigation).
2. Blekinge CoGNIT cognitive ability test (i.e., an assessment).
- Detailed Description
Worldwide, millions of people are affected by neurodegenerative diseases (e.g., Alzheimer's disease, dementia). Those diseases are having a tremendous socio-economic impact on our society. The cost associated with treating and caring for those diseases is enormous. Overwhelming evidence indicates how selective lifestyle changes (e.g., reducing exposure to known risk factors) can sometimes significantly decrease the probability of developing the disease or delay its onset. However, the diseases must be diagnosed early for them to be effective. There is a lack of accessible, inexpensive, and non-invasive practices that would allow for an early diagnosis of different diseases, even at the primary physician's office. Mantis Photonics and Blekinge Tekniska Högskola (Institustionen för Hälsa) aim to fill this urgent unmet medical need.
Strong indications of the possibility of classifying Alzheimer's status based on hyperspectral scans of the retina have been published by different researchers. These results were obtained based on images taken with hyperspectral cameras with a different working principle than the Mantis Photonics camera. The working principle of the Mantis Photonics camera allows making a hyperspectral retinoscopy with the same spectral range and comparable or better spectral resolution with a machine that is more modular and lower in cost. There is thus reason to hypothesize retinal scans taken with the Mantis Photonics camera can be used for the same classification task.
Previous studies on the automated tablet computer cognitive test CoGNIT have established validity, reliability and sensitivity for testing patients with Normal Pressure Hydrocephalus (NPH) . Recently feasibility of testing in Mild Cognitive Impairment (MCI) was affirmed (Behrens, Berglund, \& Anderberg, CoGNIT Automated Tablet Computer Cognitive Testing in Patients With Mild Cognitive Impairment: Feasibility Study, 2022). In NPH patients, CoGNIT was more sensitive to cognitive impairment at baseline and cognitive improvement after shunt surgery than the Mini-Mental State Examination (MMSE).
Blood tests for amyloid-β and other biomarkers related to Alzheimer's disease are being investigated for clinical practice, but the technique is not accepted as a standard test. Research has shown that renal function influences amyloid-β clearance from the body. Also, analytical errors influence test results. Therefore, one can question the influence of normal repeatability of the blood test result.
The aim of this investigation is the evaluation, (further) development and comparison of non-invasive techniques for the evaluation of patients suffering mild cognitive impairment, in particular, the Mantis Photonics hyperspectral camera with classification machine learning model in combination with the CoGNIT test of Dr Behrens (Blekinge Tekniska Högskola). These techniques will be compared to the result of cerebrospinal fluid analysis (CSF), the reference biological diagnostic technique for Alzheimer's disease.
Recruitment & Eligibility
- Status
- NOT_YET_RECRUITING
- Sex
- All
- Target Recruitment
- 80
- subject age over 18 years old
- The subject has undergone a lumbar puncture an cerebrospinal fluid analysis as part of the standard care.
- The subject has at least one healthy eye.
- The subject is applicable for taking a blood sample for the blood analysis test.
- The informed consent is provided, explained and understood by the person. The person has consented to the informed consent.
- There are contra-indications for lumbar puncture (eg: brain tumor with suspicion of raised intracranial pressure, coagulopathies or ongoing anticoagulant medications) will be excluded from the study.
- When the subject suffers from excessive visual or auditive impairment, the he/she will be excluded from the CoGNIT track.
Study & Design
- Study Type
- INTERVENTIONAL
- Study Design
- SINGLE_GROUP
- Arm && Interventions
Group Intervention Description Subjects non-invasive hyperspectral retinoscopy On all subjects included in the study (see inclusion / exclusion criteria and informed consent) both procedures will be performed. The result of these procedures (retinal scan, result from cognitive test and blood sample) will be used to build diagnostic classification models. Subjects blood sample On all subjects included in the study (see inclusion / exclusion criteria and informed consent) both procedures will be performed. The result of these procedures (retinal scan, result from cognitive test and blood sample) will be used to build diagnostic classification models. Subjects Test of cognitive ability on tablet computer with CoGNIT software On all subjects included in the study (see inclusion / exclusion criteria and informed consent) both procedures will be performed. The result of these procedures (retinal scan, result from cognitive test and blood sample) will be used to build diagnostic classification models.
- Primary Outcome Measures
Name Time Method Sensitivity (Statistical metric) retinal image classification model within 2 months after last patient procedure Performance metrics of the retinal image classification model: Sensitivity \[percent\]
CoGNIT test diagnostic accuracy within 2 months after last patient procedure Accuracy \[percent\] of diagnosis based on the CoGNIT test data
Accuracy (Statistical metric) retinal image classification model within 2 months after last patient procedure Performance metric of the retinal image classification model: model accuracy \[percent\]
Area under the Curve (statistical metrics) retinal image classification model within 2 months after last patient procedure Performance metric of the retinal image classification model: Area under the Curve (AuC) \[0 \< AuC \< 1\]
- Secondary Outcome Measures
Name Time Method Non invasive test variability compared to reference within 3 months after last patient procedure The variability \[relative and normalized: percent\] between the first and the second hyperspectral retinoscopy result will be compared to the variability between the blood analysis at the first and the second appointment \[relative and normalized: percent\]. The blood test variability will be used as a reference in this study.
Accuracy: Metrics combination model within 3 months after last patient procedure A combination model of both non-invasive techniques will be evaluated based on the same metrics as the single-technique model (see primary objectives) and evaluated based on the comparison of said metrics: accuracy \[percent\] for the optimal choice of threshold.
Sensitivity: Metrics combination model within 3 months after last patient procedure A combination model of both non-invasive techniques will be evaluated based on the same metrics as the single-technique model (see primary objectives) and evaluated based on the comparison of said metrics: sensitivity \[percent\] for the optimal choice of threshold.
Area Under the Curve: Metrics combination model within 3 months after last patient procedure A combination model of both non-invasive techniques will be evaluated based on the same metrics as the single-technique model (see primary objectives) and evaluated based on the comparison of said metrics: Area Under the Curve \[0\<AUC\<1\] for the optimal choice of threshold.
Trial Locations
- Locations (2)
Blekinge Tekniska Högskola
🇸🇪Karlskrona, Blekine Län, Sweden
Blekinge Hospital
🇸🇪Karlskrona, Blekinge Län, Sweden