The Application of an AI-driven Multimodal Predictive Model for Cognitive Impairment in Patients With Type 2 Diabetes Mellitus

Prior treatment

• Recent medication use (within 1 month):
  • Anti-Parkinsonian drugs, antiepileptics, sedatives/hypnotics.
• Cognitive-enhancing drugs (e.g., donepezil, memantine).

Other illness

• Other types of diabetes (e.g., type 1 diabetes, gestational diabetes).
• Metabolic disorders that may affect cognition, including:
  • Acute carbohydrate metabolism events within 3 months (e.g., severe hypoglycemia, diabetic ketoacidosis, hyperglycemic hyperosmolar state).
  • Hypothyroidism or other endocrine disorders.
• Neurological/psychiatric conditions that may impair cognition:
  • History of head trauma, depression, anxiety, delirium, severe pulmonary/renal disease, heart failure, or malignancy.

Age

• Education level \<6 years.

Other

• Substance abuse (past 2 years), including nicotine/alcohol dependence.
• Inability to complete AI screening or follow-up due to:
  • Speech/hearing/visual impairments or other disabilities

• Diagnosed with type 2 diabetes mellitus (T2DM).
• Aged 40-75 years, with self-reported or informant-reported memory complaints.
• Cognitive assessment scores consistent with either:
  • Cognitive Normal (CN) or
  • Mild Cognitive Impairment (MCI).
• Preserved ability to perform basic activities of daily living (ADLs).
• Memory-specific deficits only (e.g., partial forgetfulness), with intact other cognitive domains (e.g., reasoning, judgment, attention, and executive function).