LusogTimbang: Contactless child stunting scanner and ₱120 palengke meal optimizer for Barangay Nutrition Scholars

Problem Statement

Nearly 1 in 4 Filipino children under five years old suffers from chronic malnutrition (stunting), which permanently harms brain development and school performance. Every year during Operation Timbang Plus (OPT Plus), volunteer Barangay Nutrition Scholars (BNS) must measure the height, length, and weight of every child aged 0 to 59 months across the barangay. Measuring a squirming toddler with wooden length boards, tape measures, and paper WHO look-up tables leads to frequent transcription mistakes and months-long reporting delays. Worse, when a child is diagnosed as stunted or underweight, low-income mothers are often told to “feed more nutritious food” without a realistic plan that fits a ₱100 to ₱150 daily talipapa budget.

Proposed Solution

LusogTimbang pairs computer-vision child anthropometry with a local palengke linear-programming meal optimizer built for Barangay Nutrition Scholars and mothers:

1. Contactless Smartphone Anthropometry: The BNS places a standard laminated A4 ArUco reference card on the wall or mat beside the child and takes a 3-second photo burst. LusogTimbang extracts 3D body keypoints (crown of head, shoulders, hips, knees, heels) and upper-arm contour width, computes standing height/recumbent length and Mid-Upper Arm Circumference (MUAC) to millimeter accuracy, and calculates exact WHO Child Growth Standard Z-scores (Height-for-Age, Weight-for-Height, Weight-for-Age) offline on the phone.
2. ₱120/Day Palengke Linear-Programming Meal Optimizer: If a child falls into Stunted, Moderately Wasted, or Severely Wasted categories, LusogTimbang does not output generic textbook advice. Instead, it runs a linear-programming diet optimizer (modeled on WHO/FANTA Optifood) constrained by DOST-FNRI Philippine Food Composition Tables and local palengke prices. It outputs a 7-day, minimum-cost Filipino toddler menu (for example, ginisang monggo with tinapa flakes and malunggay, lugaw with egg and kalabasa puree, dilis powder topper) that hits 100 percent of the child’s catch-up protein, iron, zinc, calcium, and vitamin A requirements for under ₱120 per day.

Target Market

• Primary users: Barangay Nutrition Scholars (over 49,000 BNS volunteers nationwide), Rural Health Unit nutritionist-dietitians, and parents of children aged 6 to 59 months.
• Public-sector buyers: National Nutrition Council (NNC), municipal and city nutrition committees, DSWD Supplementary Feeding Program, and DOH First 1,000 Days program (RA 11148).
• Community benefit: Eliminates paper look-up errors during Operation Timbang Plus and gives low-income mothers an affordable, centavo-accurate talipapa shopping list.

How AI Is Used

• Fiducial-calibrated pose and contour anthropometry: Combines MediaPipe skeletal keypoint detection with ArUco planar homography and silhouette segmentation to measure standing stature and MUAC without bulky wooden infantometers.
• Automated WHO Z-score and growth-velocity flagging: Computes exact standard-deviation scores by age in days and biological sex, flagging faltering growth trajectories across monthly checkups.
• Linear programming nutrient optimization: Solves a constrained cost-minimization problem over DOST-FNRI nutrient matrices (energy, protein, iron, vitamin A, zinc, calcium) and local market prices to find the cheapest combination of local ingredients that closes the child’s exact micronutrient gap.
• LLM recipe customizer in local dialects: Turns the mathematically optimal gram weights into simple household takal measures (spoons, cups, tali, gatang) and cooking steps in Tagalog, Bisaya, or Ilocano.

Community & Ecosystem Integration

• Directly supports RA 11148 (Kalusugan at Nutrisyon ng Mag-Nanay Act): Equips frontline Barangay Nutrition Scholars with digital OPT Plus reporting that exports directly to the NNC e-OPT Plus spreadsheet format.
• Uses DOST-FNRI local food data: Prioritizes indigenous, backyard, and cheap wet-market ingredients (malunggay, kangkong, dilis, galunggong, monggo, kamote) instead of expensive imported supplements.
• Offline barangay operation: Both the vision height/MUAC measurement and the linear-programming diet solver run offline during remote sitio weighing visits.

Demo Scenario (3 minutes for judges)

1. Stand a child mannequin or life-size toddler cutout next to a printed ArUco card on stage.
2. Snap a photo with a phone. In 2 seconds, LusogTimbang displays the segmented keypoints, estimates height at 79.4 cm and MUAC at 12.1 cm, and computes HAZ = -2.4 SD (Stunted) and WHZ = -2.1 SD (Moderately Wasted) for a 24-month-old child.
3. Set the mother’s daily toddler food budget slider to ₱95/day in Region V (Bicol).
4. Click Optimize Palengke Basket. The linear solver runs in 100 milliseconds and displays a 7-day catch-up menu using malunggay, dilis, monggo, kalabasa, egg, and rice that achieves 100 percent of DOST-FNRI iron, protein, and vitamin A targets at ₱88/day, while exporting the official e-OPT Plus row.

30-Hour Build Plan

Suggested stack: Python FastAPI, OpenCV + MediaPipe Pose, SciPy / PuLP (linear programming), WHO anthro reference tables, Next.js PWA, Gemini API for Tagalog recipe formatting. Public data / APIs: DOST-FNRI Philippine Food Composition Tables (PhilFCT) and Pinggang Pinoy / PDRI guidelines, WHO Child Growth Standards LMS tables (0–60 months), PSA retail wet-market price averages.

Research & Literature References

1. “Accuracy of a smartphone-based 3D imaging tool for height measurement and stunting detection among children 24–59 months of age in Nepal: a validation study,” The Lancet Regional Health – Southeast Asia, 100774, 2026. DOI: 10.1016/j.lansea.2026.100774
   • How LusogTimbang uses this study: Validates our core methodology of using smartphone camera imaging to measure child stature and detect stunting in low-resource community settings as a fast, portable alternative to heavy wooden height boards.

Risks & Privacy Notes

• Child image privacy (RA 10173): All skeletal keypoint extraction runs on the BNS device; the raw photo of the child is immediately discarded after extracting height and MUAC numbers, so no images of minors are stored on servers.
• Severe Acute Malnutrition (SAM) clinical referral: Any child with MUAC < 11.5 cm, WHZ < -3 SD, or bipedal edema is immediately flagged for urgent Rural Health Unit medical referral and Ready-to-Use Therapeutic Food (RUTF) management rather than household diet changes alone.
• Allergy and breastfeeding safeguards: For infants under 6 months, the app enforces exclusive breastfeeding guidelines and never recommends solid foods or infant formula substitutes (complying with the Milk Code, EO 51).
• Strict 30-hour feasibility constraint: Monocular 2D smartphone photos without LiDAR or Time-of-Flight depth sensors struggle to hit millimeter WHO height cutoffs when toddlers squirm or tilt their knees.

AI Use Disclosure (draft)

LusogTimbang uses MediaPipe Pose and OpenCV fiducial homography to estimate child height and arm circumference, linear programming to optimize nutrient-complete meals at minimum cost, and an LLM to format recipes into local household measurements. AI coding assistants helped transcribe the DOST-FNRI nutrient table schema and build the growth chart UI. All clinical cutoffs follow WHO and DOH guidelines.

Rubric Scoring