Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disorder that causes muscle paralysis, visible fasciculations, speech difficulties (dysarthria), and tongue atrophy. 50% of ALS-affected individuals die within 30 months since diagnosis, and about 20% survive 5–10 years. In Mexico, the survival average is 68.6 months. Diagnosis of ALS is often delayed due to minimal initial symptoms and outdated diagnostic protocols, reducing the time available for symptom management. Current studies focus on using biomedical signals and machine learning (ML) to aid in diagnosing ALS. This work proposes to identify ALS through facial-symmetry analysis using simple supervised ML techniques. Using the Toronto NeuroFace dataset, key facial landmarks were extracted and converted to spherical-coordinates to minimize spatial error. Pearson correlation matrices were used to reduce the landmark numbers to 24, simplifying data processing. ML models, were evaluated using accuracy, specificity, and sensitivity metrics. Our findings underline the importance of choosing appropriate coordinates and an effective methodology for data acquisition, demonstrating that high throughput can be achieved with more accessible equipment. This work presents 66.7%, 50.0% and 87.5% in accuracy, specificity and sensitivity; respectively. This methodology has the potential to contribute to earlier and more accurate detection of ALS, facilitating a more accessible and efficient diagnosis.