File size: 9,755 Bytes
e45f24a
 
 
 
 
 
 
 
 
 
 
3a60abf
 
88e38d2
e45f24a
 
 
 
 
 
 
 
 
 
b3991da
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d05d494
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e45f24a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d05d494
615de6b
 
 
e45f24a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
import streamlit as st
import matplotlib.pyplot as plt
import matplotlib.patches as patches
from torchvision import transforms
from PIL import Image
import torch
from torchvision.models.detection import fasterrcnn_resnet50_fpn
from torchvision.models.detection.faster_rcnn import FastRCNNPredictor
import cv2
import numpy as np
from matplotlib.colors import LinearSegmentedColormap
import pytorch_lightning as pl
from tqdm import tqdm
from torchvision import models

# Function Definitions

label_names = [
    "Aortic_enlargement", "Atelectasis", "Calcification", "Cardiomegaly",
    "Consolidation", "ILD", "Infiltration", "Lung_Opacity", "Nodule/Mass",
    "Other_lesion", "Pleural_effusion", "Pleural_thickening", "Pneumothorax",
    "Pulmonary_fibrosis"
    ]

class VinDetector(pl.LightningModule):
    def __init__(self, **kwargs):
        super().__init__()

        self.model = models.detection.fasterrcnn_resnet50_fpn(pretrained=True)
        num_classes = 15
        in_features = self.model.roi_heads.box_predictor.cls_score.in_features
        self.model.roi_heads.box_predictor = FastRCNNPredictor(in_features, num_classes)
        
        self.learning_rate = 1e-3
        self.batch_size = 4

    def forward(self, x):
        return self.model(x)

    def prepare_data(self):
        
        df = pd.read_csv('../input/vinbigdata-chest-xray-abnormalities-detection/train.csv')
        df = df[df['class_id'] != 14].reset_index(drop=True)
        self.train_dataset = VBDDataset(df, '../input/vinbigdata-chest-xray-original-png/train', get_train_transform())

    def train_dataloader(self):
        
        return DataLoader(self.train_dataset, batch_size=self.batch_size, shuffle=True, pin_memory=True, num_workers=4, collate_fn=collate_fn)

    def training_step(self, batch, batch_idx):
        images, targets = batch
        targets = [{k: v for k, v in t.items()} for t in targets]
        
        loss_dict = self.model(images, targets)
        loss = sum(loss for loss in loss_dict.values())
        self.log('Loss', loss, on_step=True, on_epoch=True, prog_bar=True)
        return {"loss": loss}

    def configure_optimizers(self):
        optimizer = torch.optim.SGD(self.model.parameters(), lr=self.learning_rate, momentum=0.95, weight_decay=1e-5, nesterov=True)
        scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=6, eta_min=0, verbose=True)
        return [optimizer], [scheduler]
    

def get_train_transform():
    return A.Compose([
        A.Flip(0.5),  
        ToTensorV2(p=1.0)  
    ], bbox_params={'format': 'pascal_voc', 'label_fields': ['labels']})

def get_valid_transform():
    return A.Compose([
        ToTensorV2(p=1.0)  
    ])


def collate_fn(batch):
    return tuple(zip(*batch))
    
    

def format_prediction_string(labels, boxes, scores):
    pred_strings = []
    for j in zip(labels, scores, boxes):
        pred_strings.append("{0} {1:.4f} {2} {3} {4} {5}".format(
            j[0], j[1], j[2][0], j[2][1], j[2][2], j[2][3]))
    return " ".join(pred_strings)
    
def generate_diagnostic_report(predictions, labels, threshold=0.5):
    # Initialize an empty report string
    report = "Diagnostic Report:\n\n"
    findings_present = False
    
    # Loop through each detection
    for element in range(len(predictions['boxes'])):
        score = predictions['scores'][element].cpu().numpy()
        if score > threshold:
            label_index = predictions['labels'][element].cpu().numpy() - 1
            label_name = labels[label_index]
            report += f"- {label_name} detected with probability {score:.2f}\n"
            findings_present = True

    # If no findings above the threshold, report no significant abnormalities
    if not findings_present:
        report += "No significant abnormalities detected."

    return report

def draw_boxes_cv2(image, boxes, labels, scores, threshold=0.5, font_scale=1.0, thickness=3):
    # Define your labels and their corresponding colors
    label_names = [
        "Aortic_enlargement", "Atelectasis", "Calcification", "Cardiomegaly",
    "Consolidation", "ILD", "Infiltration", "Lung_Opacity", "Nodule/Mass",
    "Other_lesion", "Pleural_effusion", "Pleural_thickening", "Pneumothorax",
    "Pulmonary_fibrosis"
    ]
    
    label2color = [
        [59, 238, 119], [222, 21, 229], [94, 49, 164], [206, 221, 133], [117, 75, 3],
        [210, 224, 119], [211, 176, 166], [63, 7, 197], [102, 65, 77], [194, 134, 175],
        [209, 219, 50], [255, 44, 47], [89, 125, 149], [110, 27, 100]
    ]

    for i, box in enumerate(boxes):
        if scores[i] > threshold:
            # Subtract 1 from label_index to match the zero-indexed Python list
            label_index = labels[i] - 1
            label_name = label_names[label_index] if label_index < len(label_names) else "Unknown"
            color = label2color[label_index] if label_index < len(label2color) else (255, 255, 255)  # Default to white for unknown labels

            label_text = f'{label_name}: {scores[i]:.2f}'
            cv2.rectangle(image, (int(box[0]), int(box[1])), (int(box[2]), int(box[3])), color, thickness)
            cv2.putText(image, label_text, (int(box[0]), int(box[1] - 10)), cv2.FONT_HERSHEY_SIMPLEX, font_scale, color, thickness)

    return image

    # Heatmap Generation Function
def plot_image_with_colored_mask_overlay_and_original(image, predictions):
    # Assuming predictions are in the same format as Faster R-CNN outputs
    boxes = predictions['boxes'].cpu().numpy()
    scores = predictions['scores'].cpu().numpy()

    # Create a blank mask matching image size
    mask = np.zeros(image.shape[:2], dtype=np.float32)

    # Fill mask based on bounding boxes and scores
    for box, score in zip(boxes, scores):
        if score > 0.5:  # Threshold can be adjusted
            x_min, y_min, x_max, y_max = map(int, box)
            mask[y_min:y_max, x_min:x_max] += score  # Increase mask intensity with score

    # Normalize mask
    normed_mask = cv2.normalize(mask, None, alpha=0, beta=1, norm_type=cv2.NORM_MINMAX, dtype=cv2.CV_32F)

    # Create a custom colormap with transparency
    colors = [(0, 0, 0, 0), (0, 0, 1, 1), (0, 1, 0, 1), (1, 1, 0, 1), (1, 0, 0, 1)]
    cmap_name = 'doctoria_heatmap'
    custom_cmap = LinearSegmentedColormap.from_list(cmap_name, colors, N=256)

    # Apply custom colormap
    heatmap = custom_cmap(normed_mask)

    # Convert heatmap to BGR format with uint8 type
    heatmap_bgr = (heatmap[:, :, 2::-1] * 255).astype(np.uint8)

    # Overlay heatmap on original image
    overlayed_image = image.copy()
    overlayed_image[mask > 0] = overlayed_image[mask > 0] * 0.5 + heatmap_bgr[mask > 0] * 0.5

    # Plotting
    fig, axs = plt.subplots(1, 2, figsize=[12, 6])

    # Original image
    axs[0].imshow(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
    axs[0].set_title('Original Image')
    axs[0].axis('off')

    # Image with heatmap
    axs[1].imshow(cv2.cvtColor(overlayed_image, cv2.COLOR_BGR2RGB))
    axs[1].set_title('Image with Heatmap Overlay')
    axs[1].axis('off')

    # Adding colorbar
    sm = plt.cm.ScalarMappable(cmap=custom_cmap, norm=plt.Normalize(0, 1))
    sm.set_array([])
    fig.colorbar(sm, ax=axs[1], orientation='vertical', fraction=0.046, pad=0.04)

    plt.show()


# Load the model
def create_model(num_classes):
    model = fasterrcnn_resnet50_fpn(pretrained=False)
    in_features = model.roi_heads.box_predictor.cls_score.in_features
    model.roi_heads.box_predictor = FastRCNNPredictor(in_features, num_classes)
    return model


# Streamlit app title
st.title("Doctoria CXR")

# Sidebar for user input
st.sidebar.title("Upload Chest X-ray Image")

# File uploader allows user to add their own image
uploaded_file = st.sidebar.file_uploader("Upload Chest X-ray image", type=["png", "jpg", "jpeg"])

# Load the model (use your model loading function)
# Ensure the model path is correct and accessible

model = create_model(num_classes=15)
model.load_state_dict(torch.load('Doctoria CXR Thoraric Full Model.pth', map_location=torch.device('cpu')))
model.eval()

def process_image(image_path):
    # Load and transform the image
    image = Image.open(image_path).convert('RGB')
    transform = get_transform()
    image = transform(image).unsqueeze(0)

    # Perform inference
    with torch.no_grad():
        prediction = model(image)

    return prediction, image

# When the user uploads a file
if uploaded_file is not None:
    # Display the uploaded image
    st.image(uploaded_file, caption="Uploaded X-ray", use_column_width=True)
    st.write("")

    # Process the uploaded image
    prediction, image_tensor = process_image(uploaded_file)

    # Convert tensor to PIL Image for display
    image_pil = transforms.ToPILImage()(image_tensor.squeeze(0)).convert("RGB")

    # Visualization and report generation
    image_np = np.array(image_pil)
    for element in range(len(prediction[0]['boxes'])):
        box = prediction[0]['boxes'][element].cpu().numpy()
        score = prediction[0]['scores'][element].cpu().numpy()
        label_index = prediction[0]['labels'][element].cpu().numpy()
        if score > 0.5:
            draw_boxes_cv2(image_np, [box], [label_index], [score], font_scale=3)  # Increased font size

    image_pil_processed = Image.fromarray(image_np)

    # Display processed image
    st.image(image_pil_processed, caption="Processed X-ray with Abnormalities Marked", use_column_width=True)

    # Generate the diagnostic report
    report = generate_diagnostic_report(prediction[0], label_names, 0.5)
    st.write(report)

# Instructions
st.sidebar.write("Instructions:")
st.sidebar.write("1. Upload an X-ray image.")
st.sidebar.write("2. View the processed image and diagnostic report.")