In the early stages of neurodegeneration, different diseases often look identical. A patient presenting with a tremor might have idiopathic Parkinson’s disease (IPD), but they could also have Multiple System Atrophy (MSA) or Progressive Supranuclear Palsy (PSP). Similarly, early memory loss can mask the distinction between Alzheimer’s Disease (AD) and Frontotemporal Dementia (FTD). This "phenotypic convergence" creates a diagnostic bottleneck: while the symptoms overlap, the underlying pathologies—and the required treatments—are vastly different.
Recent research has shifted focus from binary classification (distinguishing "sick" from "healthy") to the more difficult task of differential diagnosis. By combining high-resolution imaging, wearable sensors, and machine learning, researchers are identifying "fingerprints" of disease that remain hidden during a standard clinical exam.
Behavioral Fingerprints: Eyes and Gait
One of the most promising avenues for differentiation lies in involuntary motor signatures. While a clinician might see a general "unsteadiness," digital sensors can decompose movement into specific metrics. Heiko Gaßner and colleagues demonstrated in The Diagnostic Scope of Sensor-Based Gait Analysis in Atypical Parkinsonism: Further Observations that wearable shoe sensors can measure gait variability to distinguish IPD from atypical parkinsonism with accuracy matching expert clinical exams. Interestingly, Victoria Sidoroff found that walking slowly acts as a "stress test" for MSA patients, magnifying gait irregularities that are less obvious at normal speeds (Characterization of gait variability in multiple system atrophy and Parkinson’s disease).
Eye movements serve as another high-resolution window into the brainstem. Sarah H. Brooks identified that slowed reflexive "prosaccades" can help separate MSA from Parkinson’s (Slowed Prosaccades and Increased Antisaccade Errors As a Potential Behavioral Biomarker of Multiple System Atrophy). Because different neurodegenerative pathologies target specific cranial nerve nuclei and cerebellar circuits, the precise "choppiness" of an eye movement—detectable even via a smartphone camera (Accurate detection of cerebellar smooth pursuit eye movement abnormalities via mobile phone video and machine learning)—can reveal which circuit is failing.
Mapping Network Topology and Metabolism
While traditional MRI looks for brain shrinkage (atrophy), newer methods examine how the brain's "wiring" and "fuel consumption" change. Adeline Su Lyn Ng showed that AD and FTD cause "reciprocal" disruptions in brain networks: AD degrades the default mode network, while FTD fragments the salience network (Distinct network topology in Alzheimer’s disease and behavioral variant frontotemporal dementia).
The integration of different imaging modalities—multi-modal fusion—is proving even more powerful. In Integrity of Neurocognitive Networks in Dementing Disorders as Measured with Simultaneous PET/Functional MRI, researchers found that metabolic connectivity (how regions consume glucose together) is significantly more accurate at distinguishing AD from FTD than traditional functional MRI. This suggests that the energy a network uses is a more sensitive marker of pathology than the blood flow signals typically measured.
Molecular and Spectral Signatures
Beyond macro-scale brain changes, researchers are looking for chemical and molecular markers. Lila Lovergne developed "spectral phenotyping," which uses infrared light to create chemical fingerprints of skin cells. This method can predict diseases like Huntington’s and Alzheimer’s long before symptoms appear, suggesting that neurodegeneration is a systemic process reflected even in non-neural tissues (An infrared spectral biomarker accurately predicts neurodegenerative disease class in the absence of overt symptoms).
Similarly, the development of "liquid biopsies" focuses on extracellular vesicles (EVs)—tiny bubbles released by brain cells into the blood. By measuring the "seeding capacity" of proteins like alpha-synuclein within these vesicles, Maria Bregendahl and colleagues are working to distinguish Dementia with Lewy Bodies from Alzheimer’s without needing a spinal tap (369 Differentiating Dementia with Lewy Bodies and Alzheimer’s Disease Using Extracellular Vesicles).
The Shift to Pattern Recognition
The common thread across these 90 papers is the move away from single-feature diagnosis toward pattern recognition. Whether it is the "top-down" vs. "bottom-up" atrophy patterns identified by the SuStaIn model (Yuya Saito in Temporal Progression Patterns of Brain Atrophy in Corticobasal Syndrome and Progressive Supranuclear Palsy Revealed by Subtype and Stage Inference (SuStaIn)) or the fusion of MRI and PET scans using Vision Transformers (MORTEZA GHAHREMANI in DiaMond: Dementia Diagnosis with Multi-Modal Vision Transformers Using MRI and PET), the goal is the same.
By treating gait, eye movements, metabolic rates, and protein concentrations as a unified high-dimensional dataset, researchers are finally beginning to untangle the overlapping symptoms of neurodegeneration. This allows for earlier intervention and ensures that patients are matched with the specific therapies their unique pathology requires.