Nano-enhancement of the machine learning assisted laser-induced breakdown spectroscopic analysis of human blood serum samples for breast cancer detection

Detecting cancer early is crucial for successful therapy and improved survival rates. In our study, we classified cancerous and non-cancerous blood serum samples using ML and nano-enhancement techniques. Both supervised (decision tree, discriminant analysis, regression, ensemble, SVM, KNN, and NN) and unsupervised (PCA) ML algorithms were employed for elemental detection and classification. The blood used for LIBS analysis was intravenous blood from patients. The serum was separated by centrifuging the blood at 4000 rpm for 15 minutes. Additionally, we applied nano-enhanced LIBS (NE-LIBS) to achieve higher accuracy. Au-nanoparticles were synthesized through laser ablation. SPR and band gap of Au-nanoparticles were measured from their UV absorption spectrum. Using micro-drops of these nanoparticles, we significantly enhanced the detection of trace elements in the LIBS signal of serum samples. Notably, the testing accuracy improved across various ML algorithms: decision tree (87.3% to 93.9%), discriminant analysis (85.5% to 99.1%), logistic regression (86.8% to 98%), naïve Bayes (85.1% to 93.9%), SVM (86.6% to 99.3%), KNN (90.1% to 98.5%), Ensemble Learning (85.5% to 98.7%), and neural networks (91.4% to 98.7%). Furthermore, our unsupervised approach by applying PCA revealed distinct clusters for both serum types. This innovative approach will assist physicians in making treatment recommendations for patients with various cancers in the early stages.