Human-centred · Interpretable · Non-invasive
HISENSE — the Human-centred Intelligent System for Early detection and monitoring of Neurodegenerative diseases via voice Signals Evaluation
A FIS 3-funded research project exploring speech signals and interpretable artificial intelligence to support earlier detection and more personalised monitoring of neurodegenerative diseases such as ALS, Parkinson’s and Alzheimer’s.
FUNDED BYThe Italian Ministry of University and Research (MUR), under the Italian Science Fund — FIS 3 Call.
The project
Small changes in speech may offer valuable clues about neurological health. Neurological disorders are among the leading causes of cognitive and physical impairment — affecting roughly one in seven people worldwide — and as populations age, the need for earlier and more objective detection only grows.
HISENSE investigates how these signals could help clinicians understand and monitor neurodegenerative diseases, bringing together artificial intelligence and clinical expertise.
The project aims to identify objective vocal biomarkers and to develop transparent and human-centred tools that clinicians can understand, trust and use.
Amyotrophic lateral sclerosis · Parkinson’s disease · Alzheimer’s disease
Human-centred. Non-invasive. Interpretable.
The challenge
Today, voice disorders — particularly dysarthria — are evaluated mainly through subjective self-report questionnaires or limited instrumental tests. Established scales such as ALSFRS-R, UPDRS and SARA lack the sensitivity for early detection, so disease status is often recognised only at later, more symptomatic stages.
The research journey
Four connected steps.
One human-centred approach.
Build an annotated, ethically governed database across diverse speech tasks, linking each recording to demographic, medical-history and clinical information — with informed consent and secure data handling.
From acoustic features such as fundamental frequency, jitter and shimmer, identify markers of dysarthria severity that could reveal impairment and shifts in neurological status.
Develop models that explain their findings — transparent, not black-box — helping clinicians assess dysarthria and understand the evidence behind each conclusion.
Study change over time from repeated recordings to support progression monitoring and more personalised, data-driven clinical decisions.
Designed around people. Accessible speech analysis and understandable insights to support clinical expertise.
Why HISENSE
Beyond single-task classification, HISENSE addresses the full path from voice to clinical insight. Three complementary dimensions work together, grounded in interpretability.
Acoustic and articulatory analysis across devices, speech tasks and recording environments — a non-invasive signal that travels with the patient.
In practiceThe smartphone becomes an assessment tool: the same task, repeated over months, tracks the trajectory no single visit can show.
Transparent, explainable and calibrated models that clinicians can understand and trust, aligned with the EU AI Act rather than black-box behaviour.
In practiceEvery output comes with the reasons behind it — a “right to explanation” clinicians can act on, not only a score.
Assessment and validation over time, connecting vocal biomarkers to clinically meaningful endpoints for real-world monitoring.
In practiceSparse, repeated assessments over a patient’s journey — not a single snapshot — inform progression and treatment choice.
Conditions studied
Several neurodegenerative conditions shape speech. HISENSE focuses on the disorders whose vocal signature can be captured, tracked and — above all — interpreted.
Flaccid dysarthria from lower-motor-neuron damage — breathy voice, hypernasality and weak articulation — that can range across spastic and mixed forms.
Because both motor pathways are affected, speech impairment may progress toward mixed dysarthria in later stages, a trajectory HISENSE is designed to track.
Hypokinetic dysarthria — reduced loudness, monopitch and a faster, more effortful speech rate linked to basal-ganglia dysfunction.
These fluctuating features can act as a remote signal of motor decline, supporting monitoring between visits.
Slowing of articulation rate and fluency, together with lexical and semantic changes that can track cognitive progression.
In AD the voice becomes a window on cognition: subtle slowing and word-finding difficulty can signal decline before other measures shift.
The team
Researchers in artificial intelligence and specialists in neurodegenerative diseases work together to connect computational research with clinical needs.
Principal Investigator · Senior Researcher
ICAR-CNR
AI, biomedical signal processing & digital health.
ProfileConnecting with me on LinkedIn
A growing research team. New researchers and PhD students are planned to be recruited and fully dedicated to HISENSE.
Publications & outcomes
Follow the scientific outputs of HISENSE, from vocal biomarkers and interpretable models to publications and research resources.
Peer-reviewed publications and conference contributions developed within the HISENSE research programme.
A multidisciplinary team created a clinically annotated speech dataset and the SAND challenge to support the development and evaluation of AI models for the early identification of ALS and the prediction of dysarthria progression.
The official report of the ICASSP 2026 SAND Grand Challenge, covering dysarthria severity classification and disease-progression prediction from speech. The challenge attracted 167 teams from 32 countries and established benchmarks for both tasks.
Earlier work that informs HISENSE. These publications precede the project.
The VOC-ALS database brings together voice recordings from people with ALS and healthy participants across multiple speech tasks.
Explore the VOC-ALS datasetAn investigation of acoustic markers and machine learning for the objective assessment of dysarthria severity in ALS.
News & events
Updates, events and opportunities to meet the HISENSE community.
Project news and upcoming events will be announced here.
Contacts
For institutional information and enquiries, visit the organisations bringing HISENSE to life.
