TexNano — real-time AI for diagnosis

We develop real-time AI tools to assist doctors in diagnosing pancreatic cancer, liver disease, and heart conditions. Our focus is on providing clinicians with reliable, instant analysis to speed up patient care.

Our Research

Contact

Amrita Institute of Medical Sciences, Kochi & Amrita Vishwa Vidyapeetham, Amritapuri · Kerala

What we do

From imaging to decisions

We work across the pipeline: better models, clearer outputs, and deployment that fits the clinic.

Diagnostic AI

ML models that support accurate, interpretable diagnosis — so clinicians can trust and act on the output.

Medical language models

Models trained on medical literature to assist with clinical reasoning and documentation.

Medical imaging

Image analysis for radiology and pathology — segmentation, classification, and quality at scale.

Clinical integration

Deployment designed for existing workflows and healthcare systems.

Research

Our research work

AI for medical imaging and diagnosis — from pancreatic cancer and fatty liver to congenital heart disease and inflammatory bowel disease.

Project

EUS-ML

Pancreatic cancer detection using Endoscopic Ultrasound (EUS).

Developed tools for pancreatic cancer detection using Endoscopic Ultrasound (EUS). The project involved Segmentation (finding the pancreas and tumor) and Classification (determining if a patient has cancer). The team utilized a dataset comprising 400+ annotated patient videos.

Cancer detection

94.12% sensitivity, 86.30% specificity, and 98.44% NPV (Negative Predictive Value).

Segmentation

70% DICE score for pancreas segmentation.

Real-time performance

EUSML-Inference App runs at ~21.28 FPS for real-time use during procedures.

Project

Fatty-Liver Diagnosis

Differentiating fatty liver from non-fatty liver using AI and widely available ultrasound.

Focused on differentiating fatty liver from non-fatty liver by teaching AI to learn Liver Steatosis and Fibrosis scores from normal ultrasound images. The objective is to replicate the expensive "Gold Standard" Fibroscan results using widely available, cheaper ultrasound machines.

Impact

The project aims to achieve accuracy similar to Fibroscan while drastically reducing the cost of testing from ₹3,000–₹10,000 to approximately ₹600–₹800 per test, expanding access to millions.

Project

VR-Heart

3D heart visualization from CT for Congenital Heart Disease (CHD) and VR surgical consultations.

Addressed the challenge surgeons face in visualizing 2D scans as 3D structures for Congenital Heart Disease (CHD). The team used AI to automate the segmentation of the heart from CT scans (reducing manual time from 2–3 hours), allowing the heart models to be loaded into Virtual Reality for surgical consultations.

Whole heart segmentation

0.9 DICE score for whole heart segmentation (bloodpool).

Chamber and vessel segmentation

0.783 DICE score for chamber and vessel segmentation.

VR & 3D Lab

A VR and 3D Lab where surgeons perform VR consultations to gain confidence for complex surgeries.

Project

Ulcerative Colitis Diagnosis

AI to assess severity of IBD / Ulcerative Colitis and reduce reliance on biopsies.

Investigating the use of AI to diagnose the severity of Inflammatory Bowel Disease (IBD) / Ulcerative Colitis to make biopsies non-required. The goal is to predict histopathology Nancy score using AI, making the process less invasive, cheaper, and faster than the traditional workflow.