Anagha Shah
Research

TechFlow

A machine-learning method that estimates menstrual flow category and viscosity from a photograph, replacing recall with measurable image features.

Health technology, 2024 to 2025

Read the paper
First page of the published TechFlow paper
First page of the IJSHRE paper.
85%
accuracy predicting flow category
100
controlled samples I prepared by hand
Oct 2025
published in IJSHRE, Volume 13, Issue 10

The need

Menstrual flow is medically useful information, but it is usually recorded from memory and vague labels.

TechFlow turns a photograph of a sanitary pad into an estimate of flow category and viscosity. The aim is a low-cost route to more objective tracking without laboratory equipment.

The model

The system preprocesses each image, isolates the pad with YOLO object detection and uses HSV color segmentation to estimate blood volume and viscosity.

  1. Capture the image
  2. Standardize and normalize
  3. Isolate the pad with object detection
  4. Segment color and geometry
  5. Return a flow and viscosity class

The dataset

No existing image database fit the problem. I prepared 100 controlled samples by hand, varying color, volume and viscosity in measurable ways. I combined them with real samples. That dataset became the base of the model.

TechFlow research sample, blurred by default
TechFlow research sample, blurred by default

These photographs show menstrual-pad test material, so they are blurred until you choose to view them.

The result

The deep-learning classifier reached about 85% accuracy in predicting flow category. Stain geometry and color intensity tracked the labeled flow and viscosity classes.

Clinical feedback

I presented the system to the Deputy Lead of Gynecology at R. N. Cooper Municipal Hospital and to the K-West Ward Health Officer of the Municipal Corporation. Their feedback pointed toward extending the work through organizations that serve adolescent girls in lower-income communities.

Videos

TechFlow Technovation video. Open on YouTube
TechFlow research presentation. Open on YouTube

Publication

Menstrual Flow Estimation Using Image-Based Machine Learning, by Anagha Shah, Divya Madhavan and Vinay Vishwakarma. International Journal of Software and Hardware Research in Engineering, Volume 13, Issue 10, October 2025, pages 39 to 48.

The project was shortlisted for the second round (Biology) of the 2025 S.T. Yau High School Science Award (Asia) and was a Technovation Girls 2025 Senior Division semifinalist.