Table 8 Challenges, solutions and future directions

Missing data

Utilize data imputation technique like mean imputation, multiple imputation by chained equations, etc

Data imbalance

Utilize resampling techniques like Synthetic Minority Over-sampling Technique, etc

Diagnostics error

Expand the use of subjective diagnostics criteria

Non-uniform longitudinal data

Data harmonization

Lack of generalizability

Develop global criteria that balance scientific rigor and practical feasibility

Exclusion of diverse populations

Encourage global collaborative efforts among researchers, clinicians, and regulatory bodies, strategic recruitment of people from culturally and linguistically diverse background

Computational burdens

Utilize efficient algorithm design, high-performance computing resources, and distributed computing platforms

Patient acceptance

Increase public awareness, ensure data transparency, security, and provide psychological support

Clinician acceptance

Offer ML training to medical students and clinicians, develop explainable AI techniques, and involve clinicians in co-design of ML tools to enhance usability and trust

Lack of interpretation for ML-dementia applications

Implement and promote explainable AI techniques like LIME and SHAP to make ML decision-making transparent

Ethical and regulatory considerations

Advocate for local and international ethical guidelines and regulatory compliance, ensure continuous monitoring post-deployment