Micro Solutions

Analytical solutions for your laboratory

RU
AI module for analysis of thyroid cytology KFBio LBC Thyroid

KFBio

AI module for analysis of thyroid cytology KFBio LBC Thyroid

AI analysis of thyroid cytology: 95% accuracy for differentiating SFN and BFN

The AI module for pre-analysis of fine-needle aspiration thyroid cytology helps the cytologist to distinguish between benign, borderline and malignant changes faster and shows explainable areas of interest on the drug.

Request a quotation

The module is designed for handheld and liquid FNA drugs, including those with sparse and overlapping cellular material, it recognizes papillary cancer, suspected malignancy, follicular neoplasms (SFN), atypia of unknown value (AUS), and benign nodes, marks nuclear grooves and matte glass nuclei, and forms Structured report in less than a minute.

According to the manufacturer, the sensitivity and accuracy of differentiation of SFN and BFN is 95%, the drug field coverage is 100%; the model is trained on more than 10,000 biopsy-confirmed slides. Local installation requires a network of 100 Mbit / s, NVIDIA 2080 Ti GPU and 32 GB of RAM on the Ubuntu 22.04 server; the recommended configuration is Gigabit network, NVIDIA A4000 and 64 GB of RAM.

Analysis opportunities

  • Optimization for handheld and liquid FNA drugs, including sparse and overlapping cells
  • Recognition of papillary cancer, suspected malignancy, SFN, AUS and benign follicular nodes
  • Isolation of nuclear grooves and matte glass nuclei
  • Structured report in less than a minute

precision

  • 95% sensitivity and 95% accuracy of differentiation between SFN and BFN
  • 100% coverage of the drug field
  • Training on more than 10,000 biopsy-confirmed slides

Infrastructure

  • Local installation: network from 100 Mbit / s, NVIDIA 2080 Ti GPU, RAM from 32 GB, Ubuntu 22.04
  • Recommended configuration: Gigabit network, NVIDIA A4000, 64 GB RAM

Diagnostics

ParameterLBC Thyroid
Identified changespapillary cancer, suspected malignancy, SFN, AUS, benign follicular nodes
Time of reportingless than 1 minute

precision

ParameterLBC Thyroid
Sensitivity (SFN/BFN)95%
precision95%
Covering the drug field100%
Training sampleover 10,000 slide

Infrastructure

ParameterLBC Thyroid
Minimum requirementsnetwork from 100 Mbit / s, NVIDIA 2080 Ti GPU, RAM from 32 GB, Ubuntu 22.04
Recommended configurationGigabit network, NVIDIA A4000 GPU, 64 GB RAM