Pathology Classification Module

The Classification module assigns detected anomalies to diagnostic categories, prioritizing findings by clinical urgency.

---

🎯 Purpose

Provide structured diagnostic labels to assist clinicians in triage and treatment planning.

---

🔍 How It Works

• Leverages deep learning models trained on large annotated datasets.
• Supports multi-class and multi-label classification.
• Outputs probability scores per pathology and severity level.

---

⚙️ Key Parameters

Parameter	Type	Description	Default
threshold	float	Minimum confidence score to report a pathology	0.5
max_labels	int	Maximum number of classifications returned per image	3
priority_mode	string	Sorting method: severity or confidence	severity

tip

Adjust threshold to filter out low-confidence labels and reduce noise.

---

📦 Output Format

The response includes:

• labels: List of pathology names.
• scores: Confidence scores (0 to 1) for each label.
• priority: Clinical urgency level (e.g., low, medium, high).

{
  "classifications": [
    {
      "label": "pneumothorax",
      "score": 0.93,
      "priority": "high"
    },
    {
      "label": "atelectasis",
      "score": 0.65,
      "priority": "medium"
    }
  ]
}

---

🛠️ Usage Example

curl -X POST "https://api.neolens.ai/v1/classification" \
  -H "Authorization: Bearer <API_KEY>" \
  -F "image=@scan_ct.png" \
  -F "threshold=0.7"

---

⚠️ Limitations

• May confuse visually similar pathologies.
• Limited by the quality and representativeness of training data.
• Priority levels are indicative; final clinical judgment required.