Artificial intelligence accurately predicted metastatic events in patients with cutaneous squamous cell carcinoma (cSCC), according to research published in the Journal of Clinical Oncology.
“Whilst rates of metastatic cSCC are low, they account for a significant proportion of skin cancer related morbidity and mortality, particularly within elderly cohorts, which poses a significant burden to healthcare services,” wrote Tom William Andrew and colleagues.
The researchers noted that current tumor staging systems do not perform well when predicting metastatic risk.

“Additionally, we lack clinically validated prognostic biomarkers—highlighting the unmet need for novel risk stratification tools to guide clinical practice and improve outcomes for patients with advanced disease,” Andrew and colleagues explained.

To address this need, the researchers trained four machine learning algorithms with the intention to optimize an AI strategy for reliable risk stratification among patients with cSCC. Using a dataset of primary cSCC registrations from the UK’s Northern Cancer Registry, the study authors analyzed 7,003 cases of histologically confirmed primary cSCC between 2010-2020. Patients underwent at least 2 years of follow-up, and the primary outcome was regional and/or distant metastases. The researchers used these data to train a Logistic Regression Trainer, a Decision Tree Classifier, a Random Forest Classifier, and a fully connected artificial neural network (ANN).

The most accurate machine learning model was the ANN, with a score of 0.94. The next most accurate model was the Logistic Regression Trainer (0.82), followed by the Random Forest Classifier (0.80) and the Decision Tree Classifier (0.71). In addition, immunosuppression was the most significant risk factor for metastases (Shapley additive explanations = 0.122).

“Significant heterogeneity in current morbidity and mortality data has limited the capacity of traditional statistical models and tumor staging systems to identify very high-risk cSSC,” Andrew and coauthors concluded. “Our findings demonstrate that machine learning algorithms can accurately predict metastatic events in cSSC populations. Further development of a model userinterface is necessary to support the development of a useful risk stratification tool to guide clinical practice.”