NON-INVASIVE ACOUSTIC BIOMARKERS FOR
DEGENERATIVE DISEASE: A CRITICAL REVIEW
AND COMPUTATIONAL ROADMAP FOR SPEECH-
BASED NEURODEGENERATION AND
VIBROARTHROGRAPHIC OSTEOARTHRITIS
DETECTION
Guthikonda Rekanth
VNR Vignana Jyothi Institute of technology
Hyderabad
Critical Gaps and Computational Roadmap for Early Detection
The Clinical Challenge
Diagnosis
• World Health Organization projects 152 million dementia cases by 2050. Current
prevalence stands at 57 million cases globally, with 10 million new diagnoses annually.
• MRI/PET Imaging scans cost $1,000–$3,000 each, making them inaccessible to most
patients in low-income countries where 80% remain undiagnosed. PET imaging exposes
patients to radioactive tracers, limiting repeat assessments for monitoring disease
progression.
• The Lumbar Puncture procedure causes severe headaches in 30% of patients, making
repeated procedures impractical for longitudinal monitoring. It remains confined to
research settings, unavailable in primary care clinics where most patients seek diagnosis.
The Clinical Challenge
Diagnosis
• World Health Organization projects 152 million dementia cases by 2050. Parkinson's
cases will double to 24 million globally. Combined, these neurodegenerative diseases
affect 80+ million people with 12 million new diagnoses annually.
• 50% of patients with mild cognitive impairement progress to Alzheimers disease
within 5 years, yet only 10% receive timely diagnosis and intervention during this
critical window when anti-amyloid therapies show maximum efficacy.
• Only 30% of patients with Parkinson’s disease receive diagnosis before stage 2. By
Stage 2, 50-70% of dopamine neurons are already lost.
How Speech Reflects Brain Pathology
• Speech production integrates motor control, cognitive-language processing, and
laryngeal function. Neurodegeneration causes distinct acoustic signatures reflecting
pathological mechanisms.
• Two Distinct Pathways:
1) Motor Dysarthria (Parkinson's, PSP, ALS): Reduced vocal range, increased
jitter/shimmer, monophasic pitch
2) Cognitive-Linguistic Decline (Alzheimer's, MCI): Increased pauses, anomia,
semantic attenuation, language planning difficulties
How Speech Reflects Brain Pathology
• Reduced Vocal Range & Expressiveness
• Increased Jitter (Frequency Variation)
• Increased Shimmer (Amplitude Variation)
• Restricted Vowel Space
ACOUSTIC FEATURES IN PARKINSON'S DISEASE
These features emerge from dopamine neuron loss in basal ganglia affecting motor
sequences
Symptoms include
How Speech Reflects Brain Pathology
• Increased Silent & Filled Pauses
• Reduced Semantic Density & Complexity
• Hesitations, false starts, incomplete utterances
• Restricted Vowel Space
ACOUSTIC FEATURES IN ALZHEIMER'S &
DEMENTIA
Reflect damage to cognitive-language centers, not motor execution areas
Symptoms include
ACOUSTIC FEATURES FOR OSTEOARTHRITIS
Healthy knee:
Well-lubricated, smooth surfaces = silent motion
Knee With Osteoarthritis:
Cartilage deterioration → rough, fibrillated surfaces →
mechanical friction during movement
PROCEDURE FOR SAMPLE ACQUISITON
Patient Arrives
↓
Audio Recording
↓
Quality Check
↓
Extract Acoustic Features
↓
AI Model Prediction
↓
Risk Score
↓
Generate Clinical Report
↓
Doctor Reviews & Makes Decision
↓
Confirm Diagnosis + Start Treatment
↓
Repeat Test in 3-6 Months
↓
Track Disease Progression
CRITICAL UNADDRESSED GAPS
THE "BINARY TRAP"
Research dominated by binary classification (Disease vs. Healthy Control)
High accuracy achieved (88-95%) but Insufficient for real clinical utility
Models can't distinguish between phenotypically similar diseases with different etiologies
They cannot quantify disease progression stage
Can't track disease trajectory over time
CRITICAL UNADDRESSED GAPS
THE "VALIDATION GAP"
Models validated against subjective clinical rating scales (UPDRS, MMSE)
This is an AI proxy validated against a human proxy
Blood biomarker panels
Real-World Data Quality Issues:
Background noise
Improper device placement or handling
Network connectivity variability
User compliance and data consistency
Transition from controlled laboratory testing → autonomous operation in home settings
CRITICAL UNADDRESSED GAPS
GENERALIZATION CRISIS & POPULATION BIAS
Language Bias:
Most models trained on English-language datasets
Poor generalization to other languages and linguistic structures
Example: Spanish Parkinson's corpus (NeuroVoz) shows different acoustic patterns
Demographic Bias:
Models trained on homogenous, small lab populations
Overfit to training data characteristics (age, accent, socioeconomic status)
Dataset Limitations:
Small, curated datasets in clinical settings
Lack of cross-language and cross-population validation
No standardized protocols for data collection
CRITICAL UNADDRESSED GAPS
VAG-SPECIFIC GAPS - STANDARDIZATION CRISIS
Lack of Standardized Data Acquisition:
Varying sensor placement (different joints, different anatomical landmarks)
Different motion protocols (flexion-extension speeds vary)
Inconsistent sampling rates and filtering procedures
Solution Needed: International consensus on VAG protocols, sensor specifications, and
signal processing standards
RECOMMENDED COMPUTATIONAL ROADMAP
Addressing Generalization & Multimodal Learning
Federated Learning
Multi-Lingual & Cross-Cultural Training
Multi-Task Learning Architecture
Differential diagnosis between diseases
Clinical Utility Through Advanced Learning Paradigms
Ordinal/Contrastive Learning for Staging
Longitudinal Regression Models
Explainable AI (XAI) Integration
Smartphone-Centric Pragmatic Fusion
RECOMMENDED COMPUTATIONAL ROADMAP
Standardization & Validation
Federated Learning
Multi-Lingual & Cross-Cultural Training
Multi-Task Learning Architecture
Differential diagnosis between diseases
Clinical Utility Through Advanced Learning Paradigms
Standardized Data Collection Protocols
Multi-Center Prospective Validation Studies Diverse populations with different age, gender, ethnicity, socioeconomic status
Regulatory Pathway(FDA or CE mark pathway)
References
1.Botelho, Catarina, et al. "Speech as a biomarker for disease detection." IEEE Access (2024).
Thies, Tabea, et al. "Automatic speech analysis combined with machine learning reliably predicts the motor state in people with Parkinson’s disease." npj Parkinson's Disease, vol. 11.1,
pp. 105, 2025.
2.König, Alexandra, et al. "Automatic speech analysis for the assessment of patients with predementia and Alzheimer's disease." Alzheimer's & Dementia: Diagnosis, Assessment &
Disease Monitoring, vol. 1.1, pp. 112-124, 2015.
3.Machrowska, Anna, et al. "Application of Recurrence Quantification Analysis in the Detection of Osteoarthritis of the Knee with the Use of Vibroarthrography." Adv. Sci. Technol. Res.
J, vol. 18, pp. 19-31, 2024.
4.Machrowska, Anna, et al. "Multi-scale analysis of knee joint acoustic signals for cartilage degeneration assessment." Sensors (Basel, Switzerland), vol. 25.3, pp. 706, 2025.
5.García, Adolfo M., et al. "Speech and language markers of neurodegeneration: a call for global equity." Brain, vol. 146.12, pp. 4870-4879, 2023.
6.Martínez-Nicolás, Israel, et al. "Ten years of research on automatic voice and speech analysis of people with Alzheimer's disease and mild cognitive impairment: a systematic review
article." Frontiers in Psychology, vol. 12, pp. 620251, 2021.
7.Karthik, R., and K. Suganthi. "Explainable AI for the diagnosis of neurodegenerative diseases: Unveiling methods, opportunities, and challenges." Computer Science Review, vol. 59,
pp. 100821, 2026.
8.Li, Renjie, et al. "Smartphone automated motor and speech analysis for early detection of Alzheimer's disease and Parkinson's disease: Validation of TapTalk across 20 different
devices." Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring, vol. 16.4, pp. e70025, 2024.
9.Vizza, Patrizia, et al. "Methodologies of speech analysis for neurodegenerative diseases evaluation." International journal of medical informatics, vol. 122, pp. 45-54, 2019.
10.Tam, Johnny, et al. "Anne Rowling Neurological Speech Corpus: clinically annotated longitudinal dataset for developing speech biomarkers in neurodegenerative disorders." Proc.
Interspeech 2025, 2025.
11.Sar, Ayan, et al. "Multi-modal deep learning framework for early detection of Parkinson’s disease using neurological and physiological data for high-fidelity diagnosis." Scientific
Reports, vol. 15.1, pp. 34835, 2025.
12.Sheikh, Shakeel A., Md Sahidullah, and Ina Kodrasi. "Overview of Automatic Speech Analysis and Technologies for Neurodegenerative Disorders: Diagnosis and Assistive
Applications." IEEE Journal of Selected Topics in Signal Processing, 2025.
13.Malekroodi, Hadi Sedigh, et al. "Speech-Based Parkinson's Detection Using Pre-Trained Self-Supervised Automatic Speech Recognition (ASR) Models and Supervised Contrastive
Learning.", 2025.
14.Karpiński, Robert, et al. "Vibroarthrography as a noninvasive screening method for early diagnosis of knee osteoarthritis: A review of current research." Applied Sciences, vol. 15.1,
pp. 279, 2024.
15.Hecker, Pascal, et al. "Voice analysis for neurological disorder recognition–a systematic review and perspective on emerging trends." Frontiers in Digital Health, vol. 4, pp. 842301,
2022.

a biomedical application with acoustic features

  • 1.
    NON-INVASIVE ACOUSTIC BIOMARKERSFOR DEGENERATIVE DISEASE: A CRITICAL REVIEW AND COMPUTATIONAL ROADMAP FOR SPEECH- BASED NEURODEGENERATION AND VIBROARTHROGRAPHIC OSTEOARTHRITIS DETECTION Guthikonda Rekanth VNR Vignana Jyothi Institute of technology Hyderabad Critical Gaps and Computational Roadmap for Early Detection
  • 2.
    The Clinical Challenge Diagnosis •World Health Organization projects 152 million dementia cases by 2050. Current prevalence stands at 57 million cases globally, with 10 million new diagnoses annually. • MRI/PET Imaging scans cost $1,000–$3,000 each, making them inaccessible to most patients in low-income countries where 80% remain undiagnosed. PET imaging exposes patients to radioactive tracers, limiting repeat assessments for monitoring disease progression. • The Lumbar Puncture procedure causes severe headaches in 30% of patients, making repeated procedures impractical for longitudinal monitoring. It remains confined to research settings, unavailable in primary care clinics where most patients seek diagnosis.
  • 3.
    The Clinical Challenge Diagnosis •World Health Organization projects 152 million dementia cases by 2050. Parkinson's cases will double to 24 million globally. Combined, these neurodegenerative diseases affect 80+ million people with 12 million new diagnoses annually. • 50% of patients with mild cognitive impairement progress to Alzheimers disease within 5 years, yet only 10% receive timely diagnosis and intervention during this critical window when anti-amyloid therapies show maximum efficacy. • Only 30% of patients with Parkinson’s disease receive diagnosis before stage 2. By Stage 2, 50-70% of dopamine neurons are already lost.
  • 4.
    How Speech ReflectsBrain Pathology • Speech production integrates motor control, cognitive-language processing, and laryngeal function. Neurodegeneration causes distinct acoustic signatures reflecting pathological mechanisms. • Two Distinct Pathways: 1) Motor Dysarthria (Parkinson's, PSP, ALS): Reduced vocal range, increased jitter/shimmer, monophasic pitch 2) Cognitive-Linguistic Decline (Alzheimer's, MCI): Increased pauses, anomia, semantic attenuation, language planning difficulties
  • 5.
    How Speech ReflectsBrain Pathology • Reduced Vocal Range & Expressiveness • Increased Jitter (Frequency Variation) • Increased Shimmer (Amplitude Variation) • Restricted Vowel Space ACOUSTIC FEATURES IN PARKINSON'S DISEASE These features emerge from dopamine neuron loss in basal ganglia affecting motor sequences Symptoms include
  • 6.
    How Speech ReflectsBrain Pathology • Increased Silent & Filled Pauses • Reduced Semantic Density & Complexity • Hesitations, false starts, incomplete utterances • Restricted Vowel Space ACOUSTIC FEATURES IN ALZHEIMER'S & DEMENTIA Reflect damage to cognitive-language centers, not motor execution areas Symptoms include
  • 7.
    ACOUSTIC FEATURES FOROSTEOARTHRITIS Healthy knee: Well-lubricated, smooth surfaces = silent motion Knee With Osteoarthritis: Cartilage deterioration → rough, fibrillated surfaces → mechanical friction during movement
  • 8.
    PROCEDURE FOR SAMPLEACQUISITON Patient Arrives ↓ Audio Recording ↓ Quality Check ↓ Extract Acoustic Features ↓ AI Model Prediction ↓ Risk Score ↓ Generate Clinical Report ↓ Doctor Reviews & Makes Decision ↓ Confirm Diagnosis + Start Treatment ↓ Repeat Test in 3-6 Months ↓ Track Disease Progression
  • 9.
    CRITICAL UNADDRESSED GAPS THE"BINARY TRAP" Research dominated by binary classification (Disease vs. Healthy Control) High accuracy achieved (88-95%) but Insufficient for real clinical utility Models can't distinguish between phenotypically similar diseases with different etiologies They cannot quantify disease progression stage Can't track disease trajectory over time
  • 10.
    CRITICAL UNADDRESSED GAPS THE"VALIDATION GAP" Models validated against subjective clinical rating scales (UPDRS, MMSE) This is an AI proxy validated against a human proxy Blood biomarker panels Real-World Data Quality Issues: Background noise Improper device placement or handling Network connectivity variability User compliance and data consistency Transition from controlled laboratory testing → autonomous operation in home settings
  • 11.
    CRITICAL UNADDRESSED GAPS GENERALIZATIONCRISIS & POPULATION BIAS Language Bias: Most models trained on English-language datasets Poor generalization to other languages and linguistic structures Example: Spanish Parkinson's corpus (NeuroVoz) shows different acoustic patterns Demographic Bias: Models trained on homogenous, small lab populations Overfit to training data characteristics (age, accent, socioeconomic status) Dataset Limitations: Small, curated datasets in clinical settings Lack of cross-language and cross-population validation No standardized protocols for data collection
  • 12.
    CRITICAL UNADDRESSED GAPS VAG-SPECIFICGAPS - STANDARDIZATION CRISIS Lack of Standardized Data Acquisition: Varying sensor placement (different joints, different anatomical landmarks) Different motion protocols (flexion-extension speeds vary) Inconsistent sampling rates and filtering procedures Solution Needed: International consensus on VAG protocols, sensor specifications, and signal processing standards
  • 13.
    RECOMMENDED COMPUTATIONAL ROADMAP AddressingGeneralization & Multimodal Learning Federated Learning Multi-Lingual & Cross-Cultural Training Multi-Task Learning Architecture Differential diagnosis between diseases Clinical Utility Through Advanced Learning Paradigms Ordinal/Contrastive Learning for Staging Longitudinal Regression Models Explainable AI (XAI) Integration Smartphone-Centric Pragmatic Fusion
  • 14.
    RECOMMENDED COMPUTATIONAL ROADMAP Standardization& Validation Federated Learning Multi-Lingual & Cross-Cultural Training Multi-Task Learning Architecture Differential diagnosis between diseases Clinical Utility Through Advanced Learning Paradigms Standardized Data Collection Protocols Multi-Center Prospective Validation Studies Diverse populations with different age, gender, ethnicity, socioeconomic status Regulatory Pathway(FDA or CE mark pathway)
  • 15.
    References 1.Botelho, Catarina, etal. "Speech as a biomarker for disease detection." IEEE Access (2024). Thies, Tabea, et al. "Automatic speech analysis combined with machine learning reliably predicts the motor state in people with Parkinson’s disease." npj Parkinson's Disease, vol. 11.1, pp. 105, 2025. 2.König, Alexandra, et al. "Automatic speech analysis for the assessment of patients with predementia and Alzheimer's disease." Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring, vol. 1.1, pp. 112-124, 2015. 3.Machrowska, Anna, et al. "Application of Recurrence Quantification Analysis in the Detection of Osteoarthritis of the Knee with the Use of Vibroarthrography." Adv. Sci. Technol. Res. J, vol. 18, pp. 19-31, 2024. 4.Machrowska, Anna, et al. "Multi-scale analysis of knee joint acoustic signals for cartilage degeneration assessment." Sensors (Basel, Switzerland), vol. 25.3, pp. 706, 2025. 5.García, Adolfo M., et al. "Speech and language markers of neurodegeneration: a call for global equity." Brain, vol. 146.12, pp. 4870-4879, 2023. 6.Martínez-Nicolás, Israel, et al. "Ten years of research on automatic voice and speech analysis of people with Alzheimer's disease and mild cognitive impairment: a systematic review article." Frontiers in Psychology, vol. 12, pp. 620251, 2021. 7.Karthik, R., and K. Suganthi. "Explainable AI for the diagnosis of neurodegenerative diseases: Unveiling methods, opportunities, and challenges." Computer Science Review, vol. 59, pp. 100821, 2026. 8.Li, Renjie, et al. "Smartphone automated motor and speech analysis for early detection of Alzheimer's disease and Parkinson's disease: Validation of TapTalk across 20 different devices." Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring, vol. 16.4, pp. e70025, 2024. 9.Vizza, Patrizia, et al. "Methodologies of speech analysis for neurodegenerative diseases evaluation." International journal of medical informatics, vol. 122, pp. 45-54, 2019. 10.Tam, Johnny, et al. "Anne Rowling Neurological Speech Corpus: clinically annotated longitudinal dataset for developing speech biomarkers in neurodegenerative disorders." Proc. Interspeech 2025, 2025. 11.Sar, Ayan, et al. "Multi-modal deep learning framework for early detection of Parkinson’s disease using neurological and physiological data for high-fidelity diagnosis." Scientific Reports, vol. 15.1, pp. 34835, 2025. 12.Sheikh, Shakeel A., Md Sahidullah, and Ina Kodrasi. "Overview of Automatic Speech Analysis and Technologies for Neurodegenerative Disorders: Diagnosis and Assistive Applications." IEEE Journal of Selected Topics in Signal Processing, 2025. 13.Malekroodi, Hadi Sedigh, et al. "Speech-Based Parkinson's Detection Using Pre-Trained Self-Supervised Automatic Speech Recognition (ASR) Models and Supervised Contrastive Learning.", 2025. 14.Karpiński, Robert, et al. "Vibroarthrography as a noninvasive screening method for early diagnosis of knee osteoarthritis: A review of current research." Applied Sciences, vol. 15.1, pp. 279, 2024. 15.Hecker, Pascal, et al. "Voice analysis for neurological disorder recognition–a systematic review and perspective on emerging trends." Frontiers in Digital Health, vol. 4, pp. 842301, 2022.