One of the leading contributors to cancer-related deaths worldwide, lung cancer shows the highest mortality rate across every demographic set. A US-based cancer survey concluded that 238,340 adults were diagnosed with lung cancer in 2023 alone. Based on the type of tissue affected, lung cancer can be widely classified into two classes: Non-small-cell lung cancer (NSCLC) and small-cell lung cancer (SCLC). Of these two subtypes, NSCLC attributes a majority of 80% to 85% of all lung cancer incidences, while 10% to 15% are SCLC. Although several treatment options are available at present, clinicians face challenges with the post-treatment prognosis of the condition. In other words, cancer recurrences even after a successful treatment or surgical procedure. To overcome these limitations, scientists experimented with the XGBoost model to predict the postoperative prognosis of NSCLC. Inspired by the recent application of AI for the early diagnosis and treatment of various conditions, predictive models are used to analyze the path of the disease post-treatment. This model proved to be a massive success.
Non-small cell lung cancer cases are traditionally treated via surgical procedures. This involves resection of the tumor mass, followed by chemotherapy and radiotherapy procedures. This final treatment step consists of a battery of tests to ensure disease-free survival. The post-surgery progression can be manually predicted prognosis on specific indicators. For example, one of the most critical prognostic criteria is the tumor stage according to TNM classification; however, the expected results are only sometimes consistent with the actual occurrence. This creates a discrepancy in decided and required treatment plans. A need for better and more accurate prognostic tools is created to formulate better treatment strategies. Due to its past success in the healthcare sector, AI predictive modeling was considered by researchers to predict the therapeutic efficacy of chemotherapy. Several factors associated with postoperative prognoses were identified for this comparative study. This included senior nutritional risk index, Glasgow prognostic score, prognostic dietary index, blood cell ratios like neutrophil and lymphocyte ratio, platelet and lymphocyte ratio, etc.
Read Also: AI-Based Tumor Removal Surgeries
The study included 1,049 participants with pathological stages I to IIIA NSCLC. Their case report documented operative procedures for cancer between January 2003 and December 2016. The electronic health record system retrieved and analyzed the clinical files and follow-up data. A fixed set of clinicopathological characteristics was analyzed for all the patients. These included age at surgery, adjuvant chemotherapy, body mass index (BMI), gender, smoking status, forced vital capacity (FVC), forced expiratory volume in one second, surgical procedure performed, and histological type of cancer. Studies were conducted on pre and postoperative hematograms, carcinoembryonic antigen (CEA), and cytokeratin-19 fragments (CYFRA).
XGBoost, a decision-free model, was selected as the algorithm for this AI prognostic model. XGBoost is advantageous as compared to other AI tools due to its ability to use missing values directly as information.
The prediction accuracy for five-year DFS, OS, and CSS was reflected by AUC values of 0.890, 0.926, and 0.960, respectively. This prediction accuracy was comparable with the accuracy levels of previous models.
Lauren Sargent is a seasoned professional in the field of business, finance media and chemical & materials, boasting more than two decades of experience in the digital arena. She has more than 10 years in B2B journalism. Her primary areas of expertise include business strategy development, operational management, product launches, fostering high-quality journalism, and adeptly leading teams to excellence. In Her role, he assumes responsibility for every facet of content oversight at the News Gazzete.