AI · Computer Vision · Microbiology

Microorganism identification
at the molecular level

An advanced microscopic image analysis system that segments, classifies, and counts microorganisms with high repeatability, full result documentation, and a presentation standard ready for sales materials and PDF documents.

The solution can be offered both as a ready-made base model and as a variant tailored to the laboratory profile, imaging standards, and the type of material analyzed by the client.

≥85%

classification accuracy

~2h

full analysis time

24/7

system access

✦ Accuracy 99.6%

app.neuralligence.pl / dashboard

Mikroorganizmy

4 684

detected objects

Avg. diameter

80.7 px

global morphology

Classification distribution

E. coli

2 216 (47.3%)

P. aeruginosa

1 132 (24.2%)

S. aureus

875 (18.7%)

C. albicans

461 (9.8%)

Dedicated option

A model built from client-provided material

It is possible to prepare and fine-tune a model based on material provided by the client, including the client’s own microorganism photos, microscopic images, and reference slides.

Adapted to the real working environment

If the client has its own image material, a model can be developed that is better matched to a specific microscope type, sample preparation method, staining, magnification, and the characteristics of the strains being examined. This option increases system usefulness wherever reproducing real laboratory conditions is critical.

training or fine-tuning the model on client images, including the client’s own microorganism photographs;
adjusting the classification scope to selected species, morphologies, or diagnostic scenarios;
comparing base model quality with a version prepared for a specific input dataset;
supporting readout standardization across operators, test series, and laboratory workstations.

Methodology

Four steps to a precise result

The analysis process runs fully automatically, from image upload to a ready report with species distribution and morphological parameters.

Image upload

The user logs into the account and uploads high-resolution images of microscopic slides.

AI segmentation

The model segments every object in the image, separating individual microorganisms from the slide background using deep learning.

ML classification

Each extracted object undergoes multiclass classification with prediction confidence expressed as a percentage.

Report and dashboard

The system generates a full quantitative and qualitative report with visualization, morphological statistics, and class distribution.

Capabilities systemu

Complete analysis in one environment

🔬

Instance segmentation

Each microorganism is treated as a separate instance — the system counts biological units rather than fields, taking morphology and cell boundaries into account.
🧬

Multiclass species classification

Identification of four key pathogen species: E. coli, P. aeruginosa, S. aureus, and C. albicans with prediction confidence provided.
📐

Morphometric analysis

Automatic calculation of geometric parameters for each object: area, perimeter, circularity, solidity, and texture entropy.
📊

Interactive results dashboard

A clear interface with a global summary, per-image analysis, and segmentation visualization with zoom for every slide.
🌐

24/7 remote access

The platform runs in the cloud — submit samples from anywhere in the world and receive results without needing to be physically present in the laboratory.

Sample gallery — slide view

4 images uploaded for analysis

✦ Galeria preparatów — system obsługuje wiele zdjęć jednocześnie, analizując każde niezależnie i łącząc wyniki w zbiorczym podsumowaniu sesji.

Technologia

Segmentacja i analiza morfometryczna

Sercem systemu jest wieloetapowy potok przetwarzania obrazu, łączący najnowocześniejsze metody widzenia komputerowego z modelami klasyfikacyjnymi uczonymi na tysiącach certyfikowanych preparatów mikrobiologicznych.

Preprocessing obrazu

Normalizacja histogramu, korekcja artefaktów optycznych i standaryzacja rozdzielczości wejściowej.

Instance segmentation

Model neuronu konwolucyjnego wykrywa i oddziela poszczególne mikroorganizmy, przypisując każdemu unikalne ID.

Deep Learning · Instance Segmentation

Ekstrakcja cech morfometrycznych

Dla każdego obiektu obliczane są parametry geometryczne i teksturowe: pole, obwód, circularity, solidity, entropy.

Klasyfikacja gatunkowa

Model EfficientNet klasyfikuje każdy mikroorganizm do jednego z gatunków z oceną pewności (confidence score).

Agregacja i wizualizacja

Results łączone są w zbiorczym raporcie z nałożonymi konturami segmentacji na oryginalne obrazy preparatów.

🔬 Wizualizacja segmentacji — wynik analizy

✓ Zweryfikowane

Microorganism segmentation visualization

Detected microorganism

Alternative class

Slide background

Morphological statistics (average values)

3483

px² area

0.75

circularity

4.89

entropy

Classification base

Detected pathogen species

The system recognizes and precisely classifies four key types of microorganisms with clinical and industrial significance.

🦠

E. coli

Gram-negative rod

Najczęściej wykrywany gatunek w analizowanych próbkach — 47,3% wykrytych mikroorganizmów. Charakterystyczna morfologia pałeczkowa, Avg. diameter 81 px.

🔬

P. aeruginosa

Gram-negative rod

Oportunistyczny patogen o wysokiej oporności. Stanowi 24,2% populacji próbek. Avg. diameter 77,9 px.

⚗️

S. aureus

Gram-positive coccus

Staphylococcus aureus — 18.7% of the population. Characteristic clusters of cocci with a small diameter (37 px).

🍄

C. albicans

Yeast-like fungus

Gatunek grzybiczny o charakterystycznej morfologii — 9,8% populacji. Największe obiekty: Avg. diameter 168 px, pole 16099 px².

Analysis quality

High accuracy and repeatability

System SDM został zaprojektowany tak, aby zapewniać wysoką classification accuracy oraz stabilną, powtarzalną analizę obrazów mikroskopowych w codziennej pracy laboratoryjnej.

Process repeatability

The automatic analytical pipeline reduces the impact of subjective operator assessment and supports readout standardization across test series.

Result stability

The model analyzes images according to a uniform scheme, supporting consistency in classification, segmentation, and reporting in the laboratory environment.

The solution was prepared for applications where high accuracy, repeatability, and clear analytical process documentation are important.

Key quality benefits

support for automatic segmentation and classification of microscopic images

High

accuracy and consistency of analysis in daily work

Consistent

repeatability of results regardless of analysis scale

Complete

clarity of reporting and process documentation

Measurable benefits

Why does SDM change the rules of the game?

Automation of microbiological analysis translates into concrete, measurable savings and operational advantage for every laboratory organization.

⚡

70%

Cost reduction

Eliminating costly and time-consuming manual procedures makes it possible to perform up to 70% more analyses within the same operating budget.

🕐

~2h

Analysis time

Complete analiza mikrobiologiczna w ~2 godziny, w porównaniu z 48–72 godzinami wymaganymi przez tradycyjne metody hodowlane.

🎯

≥85%

Accuracy

Repeatable, objective classification free from human error — every analysis is performed according to the same algorithmic protocol.

🌍

24/7

Availability

The system operates continuously in the cloud. You can request analyses from anywhere in the world, at any time, without depending on laboratory working hours.

📋

Parameters per object

For each detected microorganism, the system provides data unavailable in manual counting.

🔄

∞

Scalability

The platform scales with demand — from a few samples per day to thousands of parallel analyses without degrading performance or accuracy.

Application

A solution for the cosmetics industry

SDM is primarily used in the cosmetics industry, where fast, repeatable, and well-documented microbiological analysis supports quality control, product safety, and QA/QC processes.

🧴

Cosmetics industry

The system supports rapid verification of product and component compliance with microbiological requirements, helps standardize the assessment process, and shortens the time needed to analyze material in quality workflows.

Microorganism identificationat the molecular level