About the client
Confidential Healthcare Partner (NDA)
Services delivered

The challenge
What needed to change
A healthcare partner needed to classify medical images into diagnostic categories — cancer, non-malignant tumor, and normal — automatically, instead of relying solely on manual review for an initial triage pass.
The solution
What was delivered
We built and trained a convolutional neural network (CNN) to classify medical images into three categories, using OpenCV for preprocessing and a Keras/TensorFlow CNN architecture (three convolutional blocks with max-pooling), evaluated with a full confusion matrix and per-class precision/recall/F1 metrics rather than a single headline accuracy number.
The impact
Business outcomes
- 81.48% validation accuracy after 10 training epochs, tracked against training accuracy to catch overfitting
- Per-class precision, recall, and F1-score reported via a full classification report (cancer, non-malignant tumor, normal), not just one aggregate number
- A confusion matrix used to identify exactly which classes the model confused, not just an overall accuracy claim
More customer stories
Explore related work
Free discovery session
Planning a similar programme?
Tell us what you're building. You'll hear back from an engineer, not an inbox.
- 1We reply within one business day to set up a 30-minute call.
- 2A senior engineer — not a salesperson — walks through your problem.
- 3You get a scoped proposal with timeline and cost. No obligation.
New projects & sales
[email protected]Existing clients & support
[email protected]


