File size: 9,092 Bytes
ac8852f
08ce85d
ac8852f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
08ce85d
 
 
 
 
 
 
ac8852f
08ce85d
ac8852f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
08ce85d
ac8852f
 
 
 
 
 
08ce85d
ac8852f
 
08ce85d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ac8852f
08ce85d
 
 
ac8852f
 
08ce85d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ac8852f
 
 
 
08ce85d
 
 
 
 
 
 
 
 
 
 
 
ac8852f
 
 
 
 
 
 
 
08ce85d
ac8852f
08ce85d
ac8852f
08ce85d
 
 
ac8852f
08ce85d
 
 
ac8852f
08ce85d
ac8852f
 
 
 
 
 
 
 
08ce85d
 
 
ac8852f
08ce85d
ac8852f
 
08ce85d
ac8852f
08ce85d
ac8852f
 
 
 
 
 
 
 
 
 
 
 
 
08ce85d
ac8852f
 
 
 
08ce85d
 
ac8852f
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
import gradio as gr
from transformers import AutoImageProcessor, AutoModelForImageClassification, AutoModelForCausalLM, AutoTokenizer
from PIL import Image
import torch
from typing import Tuple, Optional, Dict, Any
from dataclasses import dataclass
import random

@dataclass
class PatientMetadata:
    age: int
    smoking_status: str
    family_history: bool
    menopause_status: str
    previous_mammogram: bool
    breast_density: str
    hormone_therapy: bool

@dataclass
class AnalysisResult:
    has_tumor: bool
    tumor_size: str
    metadata: PatientMetadata

class BreastSinogramAnalyzer:
    def __init__(self):
        """Initialize the analyzer with required models."""
        print("Initializing system...")
        self.device = "cuda" if torch.cuda.is_available() else "cpu"
        print(f"Using device: {self.device}")
        
        self._init_vision_models()
        self._init_llm()
        print("Initialization complete!")

    def _init_vision_models(self) -> None:
        """Initialize vision models for abnormality detection and size measurement."""
        print("Loading detection models...")
        self.tumor_detector = AutoModelForImageClassification.from_pretrained(
            "SIATCN/vit_tumor_classifier"
        ).to(self.device).eval()
        self.tumor_processor = AutoImageProcessor.from_pretrained("SIATCN/vit_tumor_classifier")
        
        self.size_detector = AutoModelForImageClassification.from_pretrained(
            "SIATCN/vit_tumor_radius_detection_finetuned"
        ).to(self.device).eval()
        self.size_processor = AutoImageProcessor.from_pretrained(
            "SIATCN/vit_tumor_radius_detection_finetuned"
        )

    def _init_llm(self) -> None:
        """Initialize the Qwen language model for report generation."""
        print("Loading Qwen language model...")
        self.model_name = "Qwen/QwQ-32B-Preview"
        self.model = AutoModelForCausalLM.from_pretrained(
            self.model_name,
            torch_dtype="auto",
            device_map="auto"
        )
        self.tokenizer = AutoTokenizer.from_pretrained(self.model_name)

    def _generate_synthetic_metadata(self) -> PatientMetadata:
        """Generate realistic patient metadata for breast cancer screening."""
        age = random.randint(40, 75)
        smoking_status = random.choice(["Never Smoker", "Former Smoker", "Current Smoker"])
        family_history = random.choice([True, False])
        menopause_status = "Post-menopausal" if age > 50 else "Pre-menopausal"
        previous_mammogram = random.choice([True, False])
        breast_density = random.choice(["A: Almost entirely fatty", 
                                      "B: Scattered fibroglandular",
                                      "C: Heterogeneously dense",
                                      "D: Extremely dense"])
        hormone_therapy = random.choice([True, False])

        return PatientMetadata(
            age=age,
            smoking_status=smoking_status,
            family_history=family_history,
            menopause_status=menopause_status,
            previous_mammogram=previous_mammogram,
            breast_density=breast_density,
            hormone_therapy=hormone_therapy
        )

    def _process_image(self, image: Image.Image) -> Image.Image:
        """Process input image for model consumption."""
        if image.mode != 'RGB':
            image = image.convert('RGB')
        return image.resize((224, 224))

    @torch.no_grad()
    def _analyze_image(self, image: Image.Image) -> AnalysisResult:
        """Perform abnormality detection and size measurement."""
        metadata = self._generate_synthetic_metadata()

        # Detect abnormality
        tumor_inputs = self.tumor_processor(image, return_tensors="pt").to(self.device)
        tumor_outputs = self.tumor_detector(**tumor_inputs)
        tumor_probs = tumor_outputs.logits.softmax(dim=-1)[0].cpu()
        has_tumor = tumor_probs[1] > tumor_probs[0]

        # Measure size if tumor detected
        size_inputs = self.size_processor(image, return_tensors="pt").to(self.device)
        size_outputs = self.size_detector(**size_inputs)
        size_pred = size_outputs.logits.softmax(dim=-1)[0].cpu()
        sizes = ["no-tumor", "0.5", "1.0", "1.5"]
        tumor_size = sizes[size_pred.argmax().item()]

        return AnalysisResult(has_tumor, tumor_size, metadata)

    def _generate_medical_report(self, analysis: AnalysisResult) -> str:
        """Generate a clear medical report using Qwen."""
        try:
            messages = [
                {
                    "role": "system",
                    "content": "You are a radiologist providing clear and straightforward medical reports. Focus on clarity and actionable recommendations."
                },
                {
                    "role": "user",
                    "content": f"""Generate a clear medical report for this breast imaging scan:

Scan Results:
- Finding: {'Abnormal area detected' if analysis.has_tumor else 'No abnormalities detected'}
{f'- Size of abnormal area: {analysis.tumor_size} cm' if analysis.has_tumor else ''}

Patient Information:
- Age: {analysis.metadata.age} years
- Risk factors: {', '.join([
    'family history of breast cancer' if analysis.metadata.family_history else '',
    f'{analysis.metadata.smoking_status.lower()}',
    'currently on hormone therapy' if analysis.metadata.hormone_therapy else ''
    ]).strip(', ')}

Please provide:
1. A clear interpretation of the findings
2. A specific recommendation for next steps"""
                }
            ]

            text = self.tokenizer.apply_chat_template(
                messages,
                tokenize=False,
                add_generation_prompt=True
            )
            
            model_inputs = self.tokenizer([text], return_tensors="pt").to(self.model.device)
            
            generated_ids = self.model.generate(
                **model_inputs,
                max_new_tokens=128,
                temperature=0.3,
                top_p=0.9,
                repetition_penalty=1.1,
                do_sample=True
            )
            
            generated_ids = [
                output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
            ]
            
            response = self.tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

            if len(response.split()) >= 10:
                return f"""FINDINGS AND RECOMMENDATIONS:
{response}"""
            
            return self._generate_fallback_report(analysis)

        except Exception as e:
            print(f"Error in report generation: {str(e)}")
            return self._generate_fallback_report(analysis)

    def _generate_fallback_report(self, analysis: AnalysisResult) -> str:
        """Generate a clear fallback report."""
        if analysis.has_tumor:
            return f"""FINDINGS AND RECOMMENDATIONS:

Finding: An abnormal area measuring {analysis.tumor_size} cm was detected during the scan.

Recommendation: {'An immediate follow-up with conventional mammogram and ultrasound is required.' if analysis.tumor_size in ['1.0', '1.5'] else 'A follow-up scan is recommended in 6 months.'}"""
        else:
            return """FINDINGS AND RECOMMENDATIONS:

Finding: No abnormal areas were detected during this scan.

Recommendation: Continue with routine screening as per standard guidelines."""

    def analyze(self, image: Image.Image) -> str:
        """Main analysis pipeline."""
        try:
            processed_image = self._process_image(image)
            analysis = self._analyze_image(processed_image)
            report = self._generate_medical_report(analysis)
            
            return f"""SCAN RESULTS:
{'⚠️ Abnormal area detected' if analysis.has_tumor else '✓ No abnormalities detected'}
{f'Size of abnormal area: {analysis.tumor_size} cm' if analysis.has_tumor else ''}

PATIENT INFORMATION:
• Age: {analysis.metadata.age} years
• Risk Factors: {', '.join([
    'family history of breast cancer' if analysis.metadata.family_history else '',
    analysis.metadata.smoking_status.lower(),
    'currently on hormone therapy' if analysis.metadata.hormone_therapy else '',
    ]).strip(', ')}

{report}"""
        except Exception as e:
            return f"Error during analysis: {str(e)}"

def create_interface() -> gr.Interface:
    """Create the Gradio interface."""
    analyzer = BreastSinogramAnalyzer()
    
    interface = gr.Interface(
        fn=analyzer.analyze,
        inputs=[
            gr.Image(type="pil", label="Upload Breast Image for Analysis")
        ],
        outputs=[
            gr.Textbox(label="Analysis Results", lines=20)
        ],
        title="Breast Imaging Analysis System",
        description="Upload a breast image for analysis and medical assessment.",
    )
    
    return interface

if __name__ == "__main__":
    print("Starting application...")
    interface = create_interface()
    interface.launch(debug=True, share=True)