Aug. 29, 2025
Mayo Clinic researchers have demonstrated that data-driven features extracted from F-wave responses can predict amyotrophic lateral sclerosis (ALS). As described in Brain, the findings potentially can guide efforts to improve the diagnosis of ALS and determine a person's prognosis.
Nerve conduction F-wave studies are routinely used to help diagnose neuromuscular conditions, including ALS. Efforts to create ALS biomarkers based on these studies have been challenged by the difficulties of manually interpreting waveforms, especially as ALS progresses.
The Mayo Clinic researchers used a large dataset of F-wave responses to train an artificial intelligence (AI) model to predict ALS diagnosis and patients' anticipated survival. The AI model was based on time-frequency characteristics of the F-wave responses.
"Integrating such a model into the clinical workflow could help clinicians to diagnose ALS sooner and manage treatment based on estimated survival — which might improve outcomes and the quality of life of patients," says Nathan P. Staff, M.D., Ph.D., a neurologist at Mayo Clinic in Rochester, Minnesota.
The researchers performed a retrospective analysis of F-wave responses from 46,802 patients, split into a training set (45,424 patients) and a test set (1,378 patients). In the training set, 5,329 patients had a diagnosis of motor neuron disease, of which ALS is the most common clinical diagnosis. The remaining patients were categorized as controls. In the test set, 689 patients had confirmed diagnoses of ALS and were matched with 689 controls.
Key findings:
- The model trained on F-wave data accurately classified ALS and control patients and predicted ALS survival. The wavelet model outperformed a comparable model trained with clinical annotations of F-wave waveforms.
- Factors decreasing survival were high model classification probability of ALS, older age at onset and family history of ALS.
- The model helped predict patients with bulbar symptoms who might progress to widespread ALS.
The model was trained to differentiate only people with ALS from controls. However, the researchers performed an exploratory analysis of the model's ability to differentiate between ALS and conditions with similar symptoms, including inclusion body myositis, cervical radiculopathy, lumbar radiculopathy and peripheral neuropathy. "Overall, the model does well at differentiating between these groups," Dr. Staff says.
The researchers note that their findings aren't intended to present a definitive model but to demonstrate how data-driven features extracted from the time series of the F-wave responses might improve ALS diagnosis and prognosis.
"This AI approach has broad potential to benefit ALS diagnosis prognosis and clinical trial design," Dr. Staff says. "Future iterations can expand on this approach to develop a model that includes additional mimics in the training. However, the ultimate goal isn't to have the model make the clinical diagnosis but to provide probabilities for clinicians to use to make the final diagnoses themselves."
For more information
Martinez-Thompson JM, et al. Artificial intelligence models using F-wave responses predict amyotrophic lateral sclerosis. Brain. 2025;148:2320.
Refer a patient to Mayo Clinic.