Quantifying Clinical Symptoms with Machine Learning and Sensors

This direction is merged into another direction: subjective clinical assessment (subjective clinical assessment variability). Its last version stays readable and citable here.

Version 1 of 134 papers34 conceptsfrozen 2026-10-04

Traditional clinical assessments for neurological and motor disorders often rely on "snapshot" evaluations. A clinician observes a patient performing a task—like tapping their fingers or walking a few paces—and assigns a score on an ordinal scale (e.g., 0 to 4). However, as noted across decades of research, this approach suffers from significant inter-rater variability. Two experts may look at the same tremor and provide different scores, complicating longitudinal tracking and clinical trials.

The emerging research direction focuses on replacing these subjective observations with objective, automated, and continuous measurements. By leveraging computer vision, wearable sensors, and biological markers, researchers are transforming qualitative clinical "feel" into quantitative engineering metrics.

From Visual Observation to Kinematic Precision

The core intuition behind this shift is that human eyes are poorly suited to calculate high-frequency oscillations or subtle changes in velocity, but sensors and algorithms excel at it.

For example, Amy Zhao and colleagues demonstrated in A Video-Based Method for Objectively Rating Ataxia that deep-learning-based pose estimation can extract motion features from standard video that match the accuracy of expert neurologists. Instead of a clinician guessing the severity of a movement, the system calculates the actual trajectory and variance of a limb.

This precision is further enhanced by moving from 2D video to 3D space. Kyungdo Kim and Sihan Lyu introduced the TULIP dataset, showing that 3D kinematic features significantly outperform 2D methods in predicting Parkinson’s Disease (PD) severity. By capturing movement from multiple synchronized viewpoints, the system removes the "noise" of camera angles, providing a stable, objective baseline.

Engineering Metrics for Biological Symptoms

A key breakthrough in this field is the application of control theory to human biology. In Parkinson’s Disease Rigidity: Relation to Brain Connectivity and Motor Performance, Nazanin Baradaran and her team used fMRI and machine learning to show that clinical rigidity correlates with a damping ratio. This treats the patient's limb like a physical system in an engineering model; a higher damping ratio quantitatively describes the resistance to movement that a doctor would traditionally feel by hand.

Similarly, Gert Van Dijck and colleagues used a mechatronic platform in Posterior Probability Profiles for the Automated Assessment of the Recovery of Stroke Patients to calculate the "posterior probability" that a patient’s movement belongs to a healthy control group. This shifts the question from "How does the patient look?" to "How statistically similar is this force-torque profile to a healthy baseline?"

Capturing the Unseen with Wearables and Speech

While cameras capture what is visible, wearable sensors capture what is felt. Sofija Spasojević and Tihomir Ilić utilized wireless armbands to extract "Movement Performance Indicators" (MPIs) from muscle activity and motion sensors, achieving near-perfect accuracy in distinguishing PD patients from healthy controls (Quantitative Assessment of the Arm/Hand Movements in Parkinson’s Disease Using a Wireless Armband Device).

Beyond physical movement, speech has emerged as a powerful "digital biomarker." Because neurodegenerative diseases like ALS and Parkinson's affect the fine motor control required for vocalization, researchers like Liming Wang have developed the ALS Longitudinal Speech Transformer. This model uses voice recordings to predict disease progression with 91% accuracy, providing a non-invasive way to monitor patients at home.

Bridging the Gap to Clinical Trust

For these automated systems to be adopted, they must move beyond "black-box" predictions. Recent work by T. Ehsan and Bastiaan R. Bloem emphasizes interpretability. In Interpretable and Granular Video-Based Quantification of Motor Characteristics from the Finger Tapping Test in Parkinson Disease, they extracted 13 clinically-inspired features—such as "hesitation-halts"—to explain why a model assigned a specific score.

This transparency, combined with physiological markers like salivary DJ-1 (Wenyan Kang) or serum neurofilament light chains (Zhangming Zhou), creates a multi-modal framework. By grounding algorithmic ratings in both physical motion and biological data, the research direction moves toward a future where medical assessments are as objective and reproducible as a blood pressure reading.

Papers in this version

History & citationVersion 1 · 2026-10-04

Merged into another direction · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_quantifying_clinical_symptoms_with_machi_34311831,
  title = {Quantifying Clinical Symptoms with Machine Learning and Sensors},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/quantifying-clinical-symptoms-with-machine-learning-and-sensors/rdr_34311831b52549469638fbc2dfa564f8},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_quantifying_clinical_symptoms_with_machi_34311831_v1,
  title = {Quantifying Clinical Symptoms with Machine Learning and Sensors},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/quantifying-clinical-symptoms-with-machine-learning-and-sensors/rdr_34311831b52549469638fbc2dfa564f8/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}