Background Despite advancements in treatment, laryngeal cancer survival rates remain largely unchanged due to late-stage detection. Screening and monitoring patients with laryngeal leukoplakia, a precursor to malignancy, is invasive, costly, and requires repeated visits to otolaryngology clinics, which impede timely laryngeal cancer detection. Field cancerization theory suggests that when cancer develops, subcellular changes occur proximal and distal to the site of malignancy. Elastic scattering spectroscopy (ESS), a real-time, non-invasive tool capable of differentiating tissue microarchitecture, may therefore be able to detect (and screen for) laryngeal cancer by easily measuring tissue changes in the oral cavity. This preliminary study tested whether ESS measurements obtained in the oral cavity could accurately differentiate between patients who have laryngeal cancer versus laryngeal leukoplakia. Methods Under an Institutional Review Board approved protocol, 20 patients were enrolled; 10 patients with laryngeal cancer and 10 patients with laryngeal leukoplakia. All patients were clinically assessed and categorized per standard clinical practice. Patient demographics were collected, including age, race, sex, and smoking history. ESS spectral measurements were taken in several sub-anatomical sites within patients’ oral cavities. A machine-learning algorithm was developed to classify patients based on ESS spectra obtained from patients with benign laryngeal leukoplakia and from patients with laryngeal cancer. Specificity, sensitivity, negative predictive value (NPV), hit-to-climb ratio (HCR), and area under the curve (AUC) were calculated. Additional algorithms stratified spectral data by sub-anatomic site and patients’ smoking status to further refine diagnostic capability. Results The overall algorithm had a sensitivity=77%, specificity=56%, NPV=62.5%, HCR=1.73, and AUC=0.63. Stratifying by former and active smokers, algorithm sensitivities were 82% and 80%, respectively. Analysis by sub-anatomical location showed that an algorithm based exclusively on lateral tongue spectra had an AUC of 0.76, with further stratification by former and active smokers demonstrating AUCs of 0.95 and 0.81, sensitivities of 97% and 81%, and specificities of 80% and 71%, respectively. Additionally, a mucosal lip-based algorithm showed a sensitivity of 90%, specificity of 80% and NPV of 80% in former smokers. Conclusion ESS algorithms demonstrated high sensitivity in differentiating patients with leukoplakia versus squamous cell carcinoma when subclassified by smoking status and sub-anatomical site within the oral cavity. Utilization of ESS as a real-time, non-invasive screening tool may provide prompt detection of malignancy in non-specialized settings, potentially leading to timely diagnosis, treatment, and increased survivability. A follow-up study with a larger sample size.