Title : Artificial Intelligence/Machine Learning (AI/ML) models could predict the risks for brain tumors months ahead of diagnosis using routine blood markers
Abstract:
Background: Brain tumors are the second leading cause of cancer deaths among children and young adults, with an estimated 1 million Americans living with a brain tumor. Each year, 95,000 new cases are diagnosed, of which 30% are malignant with a survival rate of only 35%. Females are one and a half times more likely than males to develop brain tumors. Meningioma and Glioblastoma account for 50% of all brain tumors. The 5-year survival rate for Glioblastoma is only 10%.
Methods: Machine learning (ML) models were trained using over 2000 sets of blood test results data from patients with and without brain tumors sourced from MIMIC-IV, a hospital-wide electronic health record (EHR) dataset from Beth Israel Deaconess Medical Center, Boston, MA. A gradient boosted model was employed with 300 trees, a maximal depth of 30 layers, and gain ratio as the criterion for attribute selection. The model's performance was evaluated using 10-fold cross-validation, demonstrating optimal results. Input parameters included age, gender, and results of routine blood markers such as complete blood counts, differential counts, comprehensive metabolic panels, and lipid panels recorded up to 6 months before brain tumor diagnosis.
Results: The model was tested in a population with a brain tumor prevalence of 1%, achieving an area under the curve (AUC) of 100% and accuracy of 99%. This yielded 100% sensitivity, 99% specificity, 83% positive predictive value, and 100% negative predictive value, with an F measure of 90% and Youden index of 0.99. Creatinine, cholesterol, white blood cells, and triglycerides predominantly contributed to identifying brain tumor risk.
Conclusion: The brain and kidneys share similar microvascular structures, making them susceptible to common pathophysiological processes. Elevated serum creatinine levels have been associated with cognitive impairment in older adults with diabetes. Additionally, brain tumors can lead to metabolic changes in the body, potentially affecting kidney function and influencing serum creatinine levels. The UK Biobank found that higher pre-diagnostic total cholesterol (TC) and high-density lipoprotein cholesterol (HDL-C) levels were associated with a higher risk of glioma in men. Brain cancer cells, particularly glioblastomas, depend heavily on cholesterol for growth and survival, subverting normal cholesterol regulation mechanisms to accumulate cholesterol. Neutrophils play a significant role in the brain tumor microenvironment, potentially influencing tumor growth, angiogenesis, and treatment resistance. The immune system, including white blood cells, plays a role in detecting and eliminating abnormal cells, including cancerous cells. The neutrophil-lymphocyte ratio (NLR) has been proposed as a biomarker for brain cancer prognosis, with higher NLR
associated with higher tumor grades and poorer outcomes in glioma patients. Dysregulation or impairment of immune surveillance mechanisms could theoretically contribute to an increased risk of tumor formation or progression. AI/ML models have identified patterns in the combined values of these markers, contributing to predictive models that may help identify the risk for brain tumors months before neurological symptoms appear, facilitating prompt diagnosis and treatment.
Audience Take Away Notes:
- Understand the significance of routine blood markers in assessing brain tumor risks
- Implement AI/ML models for identifying brain tumor risk in clinical practice
- Initiate research to study the biochemical mechanisms of these markers affecting brain tumor risk
- Utilize simple routine blood tests driven by AI/ML models to monitor high-risk patients and follow up with additional imaging studies to detect brain tumors early for intervention
