InfoBayAI/resnet18-ct-pathology-classifier

CT Scan Pathology Classification Model (ResNet18)

Model Description

This model is a deep learning–based CT scan pathology classification system built using a ResNet18 architecture and trained on medical imaging data provided by InfoBay.AI.

The training pipeline processes CT scan images from multiple organ regions, applies preprocessing and normalization, and trains a convolutional neural network to classify scans into predefined disease categories.

This approach demonstrates how structured CT imaging data can be leveraged to build efficient medical image classification systems for research and automation tasks.

---

3D Visualization of CT-Scan

---

Training Pipeline

The complete pipeline used for training is as follows:

Raw CT Images → DICOM Processing → Image Normalization → Dataset Labeling → ResNet18 Training

• Data Source: CT scan datasets (multi-organ pathology cases)
• Preprocessing: DICOM decoding, pixel normalization, resizing (224×224), RGB conversion
• Normalization: Min-max scaling to [0, 1]
• Labeling: Patient-level disease classification
• Model Training: ResNet18 for multi-class classification

---

Key Insight

This model demonstrates that even a relatively small dataset of CT images can be used to train an effective deep learning classifier.

It validates the ability of structured CT datasets to support:

• Medical image pathology classification
• Automated dataset structuring
• Preprocessing pipelines for radiology AI
• Computer vision applications in healthcare

---

Dataset Split

• Train/Test Split: 80% / 20%
• Split Strategy: Random sampling
• Number of Classes: 4

---

Training Hyperparameters

• Number of Epochs: 10
• Batch Size: 8
• Learning Rate: 1e-4
• Optimizer: Adam
• Loss Function: Cross-Entropy Loss
• Input Size: 224 × 224

---

Model Performance

The model demonstrates stable performance on internal validation data:

• Accuracy: ~90%-95% (depends on data distribution)

---

Classification Labels

Class ID	Label
0	Hydropneumothorax
1	Brain Gliosis
2	Liver Cirrhosis
3	Liver Abscess

---

Usage

Install Dependencies

pip install torch torchvision pillow numpy pydicom

import torch
import torchvision.models as models
import torch.nn as nn
import json
from PIL import Image
import numpy as np

# Load labels
with open("labels.json") as f:
    labels = json.load(f)

labels = {int(k): v for k, v in labels.items()}

# Recreate model
model = models.resnet18(pretrained=False)

model.fc = nn.Sequential(
    nn.Linear(model.fc.in_features, 256),
    nn.ReLU(),
    nn.Dropout(0.4),
    nn.Linear(256, len(labels))
)

# Load weights
model.load_state_dict(torch.load("pytorch_model.bin", map_location="cpu"))
model.eval()

# Preprocess
def preprocess(image_path):
    img = Image.open(image_path).convert("RGB")
    img = img.resize((224, 224))
    img = np.array(img) / 255.0
    img = torch.tensor(img).permute(2, 0, 1).unsqueeze(0).float()
    return img

# Predict
img = preprocess("test_image.png")

with torch.no_grad():
    output = model(img)
    probs = torch.softmax(output, dim=1)
    pred = torch.argmax(probs, 1).item()

print("Prediction:", labels[pred])

---

Considerations

This model is trained on a CT image dataset of InfoBay.AI and is intended for research and evaluation purposes only.

For access to the full dataset or enterprise licensing inquiries, please contact InfoBay.AI.

Ph: +91 8303174762
Email: [email protected]