DeCare
DeCare is an Android app for in-home caregivers looking after dementia patients. The project came out of a personal experience: a family member whose dementia went undiagnosed because the early signs looked like normal aging, and because no one in the family knew where to look for help. Indonesia has 1.2 million recorded cases, 68% of them in low-to-middle income households where care is largely informal and detection tools are out of reach. DeCare combines MMSE screening with cloud-based ML prediction, daily exercise guidance, activity reminders, curated health articles, and progress tracking. Built as a capstone for Bangkit Academy 2021.
01 /The Problem
Dementia affects over 50 million people globally, with nearly 10 million new cases every year. In Indonesia alone, 1.2 million cases were recorded in 2016, a number that doubles every 20 years. 68% of cases come from low-to-middle income families, and 61% of patients stay at home without proper treatment because professional care is too expensive. The deeper problem is awareness. Families and caregivers often read early symptoms like forgetfulness and mood swings as normal aging, so the diagnosis arrives years late, when interventions matter much less.
“My grandma went missing for three days. When a neighbor found her far from home, she couldn't remember where she lived or her family's names. That's when we knew it was dementia.”
02 /The Solution
MMSE_SCREENING
Guided Mini-Mental State Examination flow with step-by-step prompts. Scores cognitive function and flags early warning signs.
ML_PREDICTION
Screening results and patient data are sent to a cloud ML model that returns a likelihood score, so the caregiver does not have to interpret the MMSE on their own.
DAILY_CARE
Cognitive exercises, scheduled activity reminders, and a small library of caregiver-oriented articles to support the patient day to day.
PROGRESS_TRACKING
Dashboard that tracks screening scores and care activity over time. Caregivers can review progress at home and bring the same view to a doctor visit.
DeCare puts an early-detection tool directly in the caregiver's hands. The app walks them through the MMSE screening step by step, sends the results to a cloud ML model for a prediction, and then keeps showing up after the screening with daily exercises, activity reminders, and care guidance. Detection is the start of the relationship with the app, not the end of it.
03 /Technical Recap
Built natively in Kotlin with Android Jetpack components. The ML pipeline runs on Google Cloud: patient data is sent via REST to a TensorFlow model that returns a dementia probability score. Firebase handles authentication, real-time database for patient records, and Cloud Messaging for the activity reminders.
class ScreeningViewModel(
private val repository: ScreeningRepository
) : ViewModel() {
private val _result = MutableLiveData<PredictionResult>()
val result: LiveData<PredictionResult> = _result
fun submitScreening(patient: Patient, answers: List<Answer>) {
viewModelScope.launch {
val score = MMSECalculator.calculate(answers)
val prediction = repository.predict(
patientData = patient,
mmseScore = score
)
_result.value = prediction
}
}
}