Quality assurance · Production

NIRAMAI Health Analytix delivers AI-powered breast cancer detection at 90% accuracy with Google Cloud

The problem

Breast cancer is the leading cancer for women in India, with mortality rates as high as 50%, yet mammography machines cost over US$200,000 — unaffordable for most Indian hospitals — and the technique is less effective for younger women with dense breasts. Medical practitioners also struggled to interpret thermography images accurately.

First attempt

Visual interpretation of thermography had a significantly lower positive predictive value than the AI approach, and mammography — the standard alternative — was radiation-based, restricted in frequency of use, and prohibitively expensive for widespread deployment in India.

Workflow diagram · grounded in source

Thermal image capture

trigger

“A high-resolution thermal sensing device and cloud-hosted analytics solution combine to enable intelligent analysis of the thermal images”

Protocol validation via deep learning

validation

“The business uses deep learning through the TensorFlow open source library to help determine whether the images meet its protocol requirements”

Feature extraction from thermal image

ai_action

“NIRAMAI Health Analytix uses image processing models to extract over 120 features – including vascularity structures and likely lesions – from the image”

ML cancer likelihood prediction

ai_action

“a novel machine learning model is built to identify whether a person is likely to have breast cancer”

Detailed report generation

output

“The solution, in fact, generates a detailed report for review and certification by a doctor”

Doctor review and certification

human_review

“using the Firebase mobile application platform to run an application that enables doctors to view and certify images on mobile devices”

ML model retraining

feedback_loop

“NIRAMAI Health Analytix is retraining its ML models in a few months to account for any false positives and false negatives, typically coming from patients that present to tertiary cancer centers with difficult cases”

Reported outcome

NIRAMAI's AI engine achieves 90% accuracy, a 27% higher accuracy rate than mammography, a 70% higher positive predictive value than visual thermography interpretation, and is effective for 32% more patients.
The solution has been tested on over 7,500 women and scaled to 25 installations.

Reported metrics

AI diagnostic accuracy90%

Accuracy improvement vs mammography27% higher

Positive predictive value vs visual thermography interpretation70% higher

Patient coverage improvement32% more patients

Show all 8 reported metrics

AI diagnostic accuracy90%

accuracy improvement vs mammography27% higher

positive predictive value vs visual thermography interpretation70% higher

patient coverage improvement32% more patients

women screened in clinical trialsmore than 7,500

active installations25

breast cancer mortality rate in India50 percent

mammography machine costmore than US$200,000

Reported stack

TensorFlowGoogle Kubernetes EngineFirebaseGoogle Cloud Platform

Source

https://cloud.google.com/customers/niramai

Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

NIRAMAI's AI engine achieves 90% accuracy, a 27% higher accuracy rate than mammography, a 70% higher positive predictive value than visual thermography interpretation, and is effective for 32% more patients.

What tools did this team use?

TensorFlow, Google Kubernetes Engine, Firebase, Google Cloud Platform.

What results were reported?

AI diagnostic accuracy: 90%; Accuracy improvement vs mammography: 27% higher; Positive predictive value vs visual thermography interpretation: 70% higher; Patient coverage improvement: 32% more patients (source-reported, not independently verified).

What failed first in this deployment?

How is this quality assurance AI workflow structured?

Thermal image capture → Protocol validation via deep learning → Feature extraction from thermal image → ML cancer likelihood prediction → Detailed report generation → Doctor review and certification → ML model retraining.

NIRAMAI's AI engine achieves 90% accuracy, a 27% higher accuracy rate than mammography, a 70% higher positive predictive value than visual thermography interpretation, and is effective for 32% more patients.The solution has been tested on over 7,500 women and scaled to 25 installations.