Spaces:
Runtime error
Runtime error
VikramSingh178
commited on
Commit
•
e27eecd
1
Parent(s):
441421d
init
Browse files- app.py +25 -0
- brain_labels.json +6 -0
- labels.json +6 -0
- models/brain_model.pth +3 -0
- models/eye_model.pth +3 -0
- models/timm_skin_model.pth +3 -0
- models/timm_xray_model.pth +3 -0
- pages/Brain.py +156 -0
- pages/Chest.py +163 -0
- pages/Model Dashboard.py +7 -0
- pages/Skin.py +145 -0
- requirements.txt +11 -0
- skin_labels.json +10 -0
app.py
ADDED
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
from PIL import Image
|
3 |
+
|
4 |
+
|
5 |
+
# Add a title
|
6 |
+
st.set_page_config(page_title="Select Diagnosis", layout="centered")
|
7 |
+
|
8 |
+
|
9 |
+
|
10 |
+
st.title("Medical Diagnosis App")
|
11 |
+
|
12 |
+
st.markdown("")
|
13 |
+
st.markdown("<li> Currently Brain Tumors , Xrays and Skin Leison Analysis are ready for diagnosis </li>"
|
14 |
+
"<li>The Models also explain what area in the images is the cause of diagnosis </li>"
|
15 |
+
"<li>Currently the models are trained on a small dataset and will be trained on a larger dataset in the future</li>"
|
16 |
+
'<li> The Application also provides generated information on how to diagnose the disease and what should the patient do in that case</li>'
|
17 |
+
,unsafe_allow_html=True)
|
18 |
+
|
19 |
+
with st.sidebar.container():
|
20 |
+
image = Image.open("/Users/vikram/Downloads/Meditechlogo.png")
|
21 |
+
st.image(image, caption='Meditech',use_column_width=True)
|
22 |
+
|
23 |
+
|
24 |
+
|
25 |
+
|
brain_labels.json
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"0":"Glinomia",
|
3 |
+
"1": "Meningomia",
|
4 |
+
"2":"notumar",
|
5 |
+
"3": "pituary"
|
6 |
+
}
|
labels.json
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"0":"Covid19",
|
3 |
+
"1": "Normal",
|
4 |
+
"2":"Pneumonia",
|
5 |
+
"3": "Tuberculosis"
|
6 |
+
}
|
models/brain_model.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:52a7a2393337e2dedca977e4729df96b2f55579c58cbd2263922d0f8752fd866
|
3 |
+
size 79728020
|
models/eye_model.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9e83d4f41fd93bd35208cb9a1643158ee33b781ea111105fe614e2738edb3fe7
|
3 |
+
size 27238762
|
models/timm_skin_model.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a5163d00507a13ccd735162fd43dd35b16cbb901bc06d407a4deb1cef194a0e2
|
3 |
+
size 16408803
|
models/timm_xray_model.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:fe8a9cda4e25a731216bb6503f36757e9ebde7251f2c1f34c1aa3489707419c4
|
3 |
+
size 93909024
|
pages/Brain.py
ADDED
@@ -0,0 +1,156 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
from PIL import Image
|
3 |
+
import torch.nn as nn
|
4 |
+
import timm
|
5 |
+
import torch
|
6 |
+
import torchmetrics
|
7 |
+
from torchmetrics import F1Score,Recall,Accuracy
|
8 |
+
import torch.optim.lr_scheduler as lr_scheduler
|
9 |
+
import torchvision.models as models
|
10 |
+
import lightning.pytorch as pl
|
11 |
+
import torchvision
|
12 |
+
from lightning.pytorch.loggers import WandbLogger
|
13 |
+
import shap
|
14 |
+
import matplotlib.pyplot as plt
|
15 |
+
import json
|
16 |
+
from transformers import pipeline, set_seed
|
17 |
+
from transformers import BioGptTokenizer, BioGptForCausalLM
|
18 |
+
text_model = BioGptForCausalLM.from_pretrained("microsoft/biogpt")
|
19 |
+
tokenizer = BioGptTokenizer.from_pretrained("microsoft/biogpt")
|
20 |
+
labels_path = '/Users/vikram/Python/Medical Diagnosis App/brain_labels.json'
|
21 |
+
from captum.attr import DeepLift , visualization
|
22 |
+
|
23 |
+
with open(labels_path) as json_data:
|
24 |
+
idx_to_labels = json.load(json_data)
|
25 |
+
|
26 |
+
|
27 |
+
|
28 |
+
class FineTuneModel(pl.LightningModule):
|
29 |
+
def __init__(self, model_name, num_classes, learning_rate, dropout_rate,beta1,beta2,eps):
|
30 |
+
super().__init__()
|
31 |
+
self.model_name = model_name
|
32 |
+
self.num_classes = num_classes
|
33 |
+
self.learning_rate = learning_rate
|
34 |
+
self.beta1 = beta1
|
35 |
+
self.beta2 = beta2
|
36 |
+
self.eps = eps
|
37 |
+
self.dropout_rate = dropout_rate
|
38 |
+
self.model = timm.create_model(self.model_name, pretrained=True,num_classes=self.num_classes)
|
39 |
+
self.loss_fn = nn.CrossEntropyLoss()
|
40 |
+
self.f1 = F1Score(task='multiclass', num_classes=self.num_classes)
|
41 |
+
self.recall = Recall(task='multiclass', num_classes=self.num_classes)
|
42 |
+
self.accuracy = Accuracy(task='multiclass', num_classes=self.num_classes)
|
43 |
+
|
44 |
+
#for param in self.model.parameters():
|
45 |
+
#param.requires_grad = True
|
46 |
+
#self.model.classifier= nn.Sequential(nn.Dropout(p=self.dropout_rate),nn.Linear(self.model.classifier.in_features, self.num_classes))
|
47 |
+
#self.model.classifier.requires_grad = True
|
48 |
+
|
49 |
+
|
50 |
+
def forward(self, x):
|
51 |
+
return self.model(x)
|
52 |
+
|
53 |
+
def training_step(self, batch, batch_idx):
|
54 |
+
x, y = batch
|
55 |
+
y_hat = self.model(x)
|
56 |
+
loss = self.loss_fn(y_hat, y)
|
57 |
+
acc = self.accuracy(y_hat.argmax(dim=1),y)
|
58 |
+
f1 = self.f1(y_hat.argmax(dim=1),y)
|
59 |
+
recall = self.recall(y_hat.argmax(dim=1),y)
|
60 |
+
self.log('train_loss', loss,on_step=False,on_epoch=True)
|
61 |
+
self.log('train_acc', acc,on_step=False,on_epoch = True)
|
62 |
+
self.log('train_f1',f1,on_step=False,on_epoch=True)
|
63 |
+
self.log('train_recall',recall,on_step=False,on_epoch=True)
|
64 |
+
return loss
|
65 |
+
|
66 |
+
def validation_step(self, batch, batch_idx):
|
67 |
+
x, y = batch
|
68 |
+
y_hat = self.model(x)
|
69 |
+
loss = self.loss_fn(y_hat, y)
|
70 |
+
acc = self.accuracy(y_hat.argmax(dim=1),y)
|
71 |
+
f1 = self.f1(y_hat.argmax(dim=1),y)
|
72 |
+
recall = self.recall(y_hat.argmax(dim=1),y)
|
73 |
+
self.log('val_loss', loss,on_step=False,on_epoch=True)
|
74 |
+
self.log('val_acc', acc,on_step=False,on_epoch=True)
|
75 |
+
self.log('val_f1',f1,on_step=False,on_epoch=True)
|
76 |
+
self.log('val_recall',recall,on_step=False,on_epoch=True)
|
77 |
+
|
78 |
+
|
79 |
+
def configure_optimizers(self):
|
80 |
+
optimizer = torch.optim.Adam(self.model.parameters(), lr=self.learning_rate,betas=(self.beta1,self.beta2),eps=self.eps)
|
81 |
+
scheduler = lr_scheduler.StepLR(optimizer, step_size=10, gamma=0.1)
|
82 |
+
return {'optimizer': optimizer, 'lr_scheduler': scheduler}
|
83 |
+
|
84 |
+
|
85 |
+
#load model
|
86 |
+
|
87 |
+
|
88 |
+
|
89 |
+
|
90 |
+
|
91 |
+
st.markdown("<h1 style='text-align: center; '>Brain Tumor Diagnosis</h1>",unsafe_allow_html=True)
|
92 |
+
|
93 |
+
|
94 |
+
|
95 |
+
|
96 |
+
# Display a file uploader widget for the user to upload an image
|
97 |
+
|
98 |
+
uploaded_file = st.file_uploader("Choose an Brain MRI image file", type=["jpg", "jpeg", "png"])
|
99 |
+
|
100 |
+
# Load the uploaded image, or display emojis if no file was uploaded
|
101 |
+
with st.container():
|
102 |
+
if uploaded_file is not None:
|
103 |
+
|
104 |
+
image = Image.open(uploaded_file)
|
105 |
+
st.image(image, caption='Diagnosis', use_column_width=True)
|
106 |
+
model = timm.create_model(model_name='efficientnet_b1', pretrained=True,num_classes=4)
|
107 |
+
data_cfg = timm.data.resolve_data_config(model.pretrained_cfg)
|
108 |
+
transform = timm.data.create_transform(**data_cfg)
|
109 |
+
model_transforms = torchvision.transforms.Compose([transform])
|
110 |
+
transformed_image = model_transforms(image)
|
111 |
+
brain_model = torch.load('models/brain_model.pth')
|
112 |
+
|
113 |
+
brain_model.eval()
|
114 |
+
with torch.inference_mode():
|
115 |
+
with st.progress(100):
|
116 |
+
|
117 |
+
#class_names = ['Glinomia','Meningomia','notumar','pituary']
|
118 |
+
prediction = torch.nn.functional.softmax(brain_model(transformed_image.unsqueeze(dim=0))[0], dim=0)
|
119 |
+
prediction_score, pred_label_idx = torch.topk(prediction, 1)
|
120 |
+
pred_label_idx.squeeze_()
|
121 |
+
predicted_label = idx_to_labels[str(pred_label_idx.item())]
|
122 |
+
st.write( f'Predicted Label: {predicted_label}')
|
123 |
+
if st.button('Know More'):
|
124 |
+
generator = pipeline("text-generation",model=text_model,tokenizer=tokenizer)
|
125 |
+
input_text = f"Patient has {predicted_label} and is advised to take the following medicines:"
|
126 |
+
with st.spinner('Generating Text'):
|
127 |
+
generator(input_text, max_length=300, do_sample=True, top_k=50, top_p=0.95, num_return_sequences=1)
|
128 |
+
st.markdown(generator(input_text, max_length=300, do_sample=True, top_k=50, top_p=0.95, num_return_sequences=1)[0]['generated_text'])
|
129 |
+
|
130 |
+
|
131 |
+
|
132 |
+
|
133 |
+
|
134 |
+
|
135 |
+
|
136 |
+
|
137 |
+
|
138 |
+
|
139 |
+
|
140 |
+
|
141 |
+
|
142 |
+
else:
|
143 |
+
st.success("Please upload an image file 🧠")
|
144 |
+
|
145 |
+
|
146 |
+
|
147 |
+
|
148 |
+
|
149 |
+
## Model Explainibilty Dashboard using Captum
|
150 |
+
|
151 |
+
|
152 |
+
|
153 |
+
|
154 |
+
|
155 |
+
|
156 |
+
|
pages/Chest.py
ADDED
@@ -0,0 +1,163 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
from PIL import Image
|
3 |
+
import torch.nn as nn
|
4 |
+
import timm
|
5 |
+
import torch
|
6 |
+
import time
|
7 |
+
import torchmetrics
|
8 |
+
from torchmetrics import F1Score,Recall,Accuracy
|
9 |
+
import torch.optim.lr_scheduler as lr_scheduler
|
10 |
+
import torchvision.models as models
|
11 |
+
import lightning.pytorch as pl
|
12 |
+
import torchvision
|
13 |
+
from lightning.pytorch.loggers import WandbLogger
|
14 |
+
import captum
|
15 |
+
import matplotlib.pyplot as plt
|
16 |
+
import json
|
17 |
+
from transformers import pipeline, set_seed
|
18 |
+
from transformers import BioGptTokenizer, BioGptForCausalLM
|
19 |
+
text_model = BioGptForCausalLM.from_pretrained("microsoft/biogpt")
|
20 |
+
tokenizer = BioGptTokenizer.from_pretrained("microsoft/biogpt")
|
21 |
+
labels_path = '/Users/vikram/Python/Medical Diagnosis App/labels.json'
|
22 |
+
|
23 |
+
|
24 |
+
with open(labels_path) as json_data:
|
25 |
+
idx_to_labels = json.load(json_data)
|
26 |
+
|
27 |
+
|
28 |
+
|
29 |
+
class FineTuneModel(pl.LightningModule):
|
30 |
+
def __init__(self, model_name, num_classes, learning_rate, dropout_rate,beta1,beta2,eps):
|
31 |
+
super().__init__()
|
32 |
+
self.model_name = model_name
|
33 |
+
self.num_classes = num_classes
|
34 |
+
self.learning_rate = learning_rate
|
35 |
+
self.beta1 = beta1
|
36 |
+
self.beta2 = beta2
|
37 |
+
self.eps = eps
|
38 |
+
self.dropout_rate = dropout_rate
|
39 |
+
self.model = timm.create_model(self.model_name, pretrained=True,num_classes=self.num_classes)
|
40 |
+
self.loss_fn = nn.CrossEntropyLoss()
|
41 |
+
self.f1 = F1Score(task='multiclass', num_classes=self.num_classes)
|
42 |
+
self.recall = Recall(task='multiclass', num_classes=self.num_classes)
|
43 |
+
self.accuracy = Accuracy(task='multiclass', num_classes=self.num_classes)
|
44 |
+
|
45 |
+
#for param in self.model.parameters():
|
46 |
+
#param.requires_grad = True
|
47 |
+
#self.model.classifier= nn.Sequential(nn.Dropout(p=self.dropout_rate),nn.Linear(self.model.classifier.in_features, self.num_classes))
|
48 |
+
#self.model.classifier.requires_grad = True
|
49 |
+
|
50 |
+
|
51 |
+
def forward(self, x):
|
52 |
+
return self.model(x)
|
53 |
+
|
54 |
+
def training_step(self, batch, batch_idx):
|
55 |
+
x, y = batch
|
56 |
+
y_hat = self.model(x)
|
57 |
+
loss = self.loss_fn(y_hat, y)
|
58 |
+
acc = self.accuracy(y_hat.argmax(dim=1),y)
|
59 |
+
f1 = self.f1(y_hat.argmax(dim=1),y)
|
60 |
+
recall = self.recall(y_hat.argmax(dim=1),y)
|
61 |
+
self.log('train_loss', loss,on_step=False,on_epoch=True)
|
62 |
+
self.log('train_acc', acc,on_step=False,on_epoch = True)
|
63 |
+
self.log('train_f1',f1,on_step=False,on_epoch=True)
|
64 |
+
self.log('train_recall',recall,on_step=False,on_epoch=True)
|
65 |
+
return loss
|
66 |
+
|
67 |
+
def validation_step(self, batch, batch_idx):
|
68 |
+
x, y = batch
|
69 |
+
y_hat = self.model(x)
|
70 |
+
loss = self.loss_fn(y_hat, y)
|
71 |
+
acc = self.accuracy(y_hat.argmax(dim=1),y)
|
72 |
+
f1 = self.f1(y_hat.argmax(dim=1),y)
|
73 |
+
recall = self.recall(y_hat.argmax(dim=1),y)
|
74 |
+
self.log('val_loss', loss,on_step=False,on_epoch=True)
|
75 |
+
self.log('val_acc', acc,on_step=False,on_epoch=True)
|
76 |
+
self.log('val_f1',f1,on_step=False,on_epoch=True)
|
77 |
+
self.log('val_recall',recall,on_step=False,on_epoch=True)
|
78 |
+
|
79 |
+
|
80 |
+
def configure_optimizers(self):
|
81 |
+
optimizer = torch.optim.Adam(self.model.parameters(), lr=self.learning_rate,betas=(self.beta1,self.beta2),eps=self.eps)
|
82 |
+
scheduler = lr_scheduler.StepLR(optimizer, step_size=10, gamma=0.1)
|
83 |
+
return {'optimizer': optimizer, 'lr_scheduler': scheduler}
|
84 |
+
|
85 |
+
|
86 |
+
#load model
|
87 |
+
|
88 |
+
|
89 |
+
|
90 |
+
|
91 |
+
|
92 |
+
st.markdown("<h1 style='text-align: center; '>Chest Xray Diagnosis</h1>",unsafe_allow_html=True)
|
93 |
+
|
94 |
+
|
95 |
+
|
96 |
+
|
97 |
+
# Display a file uploader widget for the user to upload an image
|
98 |
+
uploaded_file = st.file_uploader("Choose an Chest XRay Image file", type=["jpg", "jpeg", "png"])
|
99 |
+
|
100 |
+
# Load the uploaded image, or display emojis if no file was uploaded
|
101 |
+
if uploaded_file is not None:
|
102 |
+
|
103 |
+
image = Image.open(uploaded_file)
|
104 |
+
st.image(image, caption='Diagnosis',width=224, use_column_width=True)
|
105 |
+
model = timm.create_model(model_name='efficientnet_b2', pretrained=True,num_classes=4)
|
106 |
+
data_cfg = timm.data.resolve_data_config(model.pretrained_cfg)
|
107 |
+
transform = timm.data.create_transform(**data_cfg)
|
108 |
+
model_transforms = torchvision.transforms.Compose([transform])
|
109 |
+
transformed_image = model_transforms(image)
|
110 |
+
xray_model = torch.load('models/timm_xray_model.pth')
|
111 |
+
|
112 |
+
xray_model.eval()
|
113 |
+
|
114 |
+
|
115 |
+
|
116 |
+
with torch.inference_mode():
|
117 |
+
with st.progress(100):
|
118 |
+
|
119 |
+
prediction = torch.nn.functional.softmax(xray_model(transformed_image.unsqueeze(dim=0))[0], dim=0)
|
120 |
+
prediction_score, pred_label_idx = torch.topk(prediction, 1)
|
121 |
+
pred_label_idx.squeeze_()
|
122 |
+
predicted_label = idx_to_labels[str(pred_label_idx.item())]
|
123 |
+
st.write( f'Predicted Label: {predicted_label}')
|
124 |
+
if st.button('Know More'):
|
125 |
+
generator = pipeline("text-generation",model=text_model,tokenizer=tokenizer)
|
126 |
+
input_text = f"Patient has {predicted_label} and is advised to take the following medicines:"
|
127 |
+
with st.spinner('Generating Text'):
|
128 |
+
generator(input_text, max_length=300, do_sample=True, top_k=50, top_p=0.95, num_return_sequences=1)
|
129 |
+
st.markdown(generator(input_text, max_length=300, do_sample=True, top_k=50, top_p=0.95, num_return_sequences=1)[0]['generated_text'])
|
130 |
+
|
131 |
+
|
132 |
+
|
133 |
+
|
134 |
+
|
135 |
+
|
136 |
+
|
137 |
+
|
138 |
+
|
139 |
+
|
140 |
+
|
141 |
+
else:
|
142 |
+
st.success("Please upload an image file ⚕️")
|
143 |
+
|
144 |
+
|
145 |
+
|
146 |
+
|
147 |
+
|
148 |
+
|
149 |
+
|
150 |
+
|
151 |
+
|
152 |
+
|
153 |
+
|
154 |
+
|
155 |
+
|
156 |
+
|
157 |
+
|
158 |
+
|
159 |
+
|
160 |
+
|
161 |
+
|
162 |
+
|
163 |
+
|
pages/Model Dashboard.py
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
|
3 |
+
st.components.v1.iframe(src = 'https://api.wandb.ai/links/vikramxd/nw5ru81j',width = 1000, height = 800,scrolling = True)
|
4 |
+
|
5 |
+
|
6 |
+
|
7 |
+
|
pages/Skin.py
ADDED
@@ -0,0 +1,145 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
from PIL import Image
|
3 |
+
import torch.nn as nn
|
4 |
+
import timm
|
5 |
+
import torch
|
6 |
+
import torchmetrics
|
7 |
+
from torchmetrics import F1Score,Recall,Accuracy
|
8 |
+
import torch.optim.lr_scheduler as lr_scheduler
|
9 |
+
import torchvision.models as models
|
10 |
+
import lightning.pytorch as pl
|
11 |
+
import torchvision
|
12 |
+
from lightning.pytorch.loggers import WandbLogger
|
13 |
+
import shap
|
14 |
+
import matplotlib.pyplot as plt
|
15 |
+
import json
|
16 |
+
from transformers import pipeline, set_seed
|
17 |
+
from transformers import BioGptTokenizer, BioGptForCausalLM
|
18 |
+
text_model = BioGptForCausalLM.from_pretrained("microsoft/biogpt")
|
19 |
+
tokenizer = BioGptTokenizer.from_pretrained("microsoft/biogpt")
|
20 |
+
labels_path = '/Users/vikram/Python/Medical Diagnosis App/skin_labels.json'
|
21 |
+
from captum.attr import DeepLift , visualization
|
22 |
+
|
23 |
+
with open(labels_path) as json_data:
|
24 |
+
idx_to_labels = json.load(json_data)
|
25 |
+
|
26 |
+
|
27 |
+
|
28 |
+
class FineTuneModel(pl.LightningModule):
|
29 |
+
def __init__(self, model_name, num_classes, learning_rate, dropout_rate,beta1,beta2,eps):
|
30 |
+
super().__init__()
|
31 |
+
self.model_name = model_name
|
32 |
+
self.num_classes = num_classes
|
33 |
+
self.learning_rate = learning_rate
|
34 |
+
self.beta1 = beta1
|
35 |
+
self.beta2 = beta2
|
36 |
+
self.eps = eps
|
37 |
+
self.dropout_rate = dropout_rate
|
38 |
+
self.model = timm.create_model(self.model_name, pretrained=True,num_classes=self.num_classes)
|
39 |
+
self.loss_fn = nn.CrossEntropyLoss()
|
40 |
+
self.f1 = F1Score(task='multiclass', num_classes=self.num_classes)
|
41 |
+
self.recall = Recall(task='multiclass', num_classes=self.num_classes)
|
42 |
+
self.accuracy = Accuracy(task='multiclass', num_classes=self.num_classes)
|
43 |
+
|
44 |
+
#for param in self.model.parameters():
|
45 |
+
#param.requires_grad = True
|
46 |
+
#self.model.classifier= nn.Sequential(nn.Dropout(p=self.dropout_rate),nn.Linear(self.model.classifier.in_features, self.num_classes))
|
47 |
+
#self.model.classifier.requires_grad = True
|
48 |
+
|
49 |
+
|
50 |
+
def forward(self, x):
|
51 |
+
return self.model(x)
|
52 |
+
|
53 |
+
def training_step(self, batch, batch_idx):
|
54 |
+
x, y = batch
|
55 |
+
y_hat = self.model(x)
|
56 |
+
loss = self.loss_fn(y_hat, y)
|
57 |
+
acc = self.accuracy(y_hat.argmax(dim=1),y)
|
58 |
+
f1 = self.f1(y_hat.argmax(dim=1),y)
|
59 |
+
recall = self.recall(y_hat.argmax(dim=1),y)
|
60 |
+
self.log('train_loss', loss,on_step=False,on_epoch=True)
|
61 |
+
self.log('train_acc', acc,on_step=False,on_epoch = True)
|
62 |
+
self.log('train_f1',f1,on_step=False,on_epoch=True)
|
63 |
+
self.log('train_recall',recall,on_step=False,on_epoch=True)
|
64 |
+
return loss
|
65 |
+
|
66 |
+
def validation_step(self, batch, batch_idx):
|
67 |
+
x, y = batch
|
68 |
+
y_hat = self.model(x)
|
69 |
+
loss = self.loss_fn(y_hat, y)
|
70 |
+
acc = self.accuracy(y_hat.argmax(dim=1),y)
|
71 |
+
f1 = self.f1(y_hat.argmax(dim=1),y)
|
72 |
+
recall = self.recall(y_hat.argmax(dim=1),y)
|
73 |
+
self.log('val_loss', loss,on_step=False,on_epoch=True)
|
74 |
+
self.log('val_acc', acc,on_step=False,on_epoch=True)
|
75 |
+
self.log('val_f1',f1,on_step=False,on_epoch=True)
|
76 |
+
self.log('val_recall',recall,on_step=False,on_epoch=True)
|
77 |
+
|
78 |
+
|
79 |
+
def configure_optimizers(self):
|
80 |
+
optimizer = torch.optim.Adam(self.model.parameters(), lr=self.learning_rate,betas=(self.beta1,self.beta2),eps=self.eps)
|
81 |
+
scheduler = lr_scheduler.StepLR(optimizer, step_size=10, gamma=0.1)
|
82 |
+
return {'optimizer': optimizer, 'lr_scheduler': scheduler}
|
83 |
+
|
84 |
+
|
85 |
+
#load model
|
86 |
+
|
87 |
+
|
88 |
+
|
89 |
+
|
90 |
+
|
91 |
+
st.markdown("<h1 style='text-align: center; '>Skin Leision Diagnosis</h1>",unsafe_allow_html=True)
|
92 |
+
|
93 |
+
|
94 |
+
|
95 |
+
|
96 |
+
# Display a file uploader widget for the user to upload an image
|
97 |
+
|
98 |
+
uploaded_file = st.file_uploader("Choose an Skin image file", type=["jpg", "jpeg", "png"])
|
99 |
+
|
100 |
+
# Load the uploaded image, or display emojis if no file was uploaded
|
101 |
+
with st.container():
|
102 |
+
if uploaded_file is not None:
|
103 |
+
|
104 |
+
image = Image.open(uploaded_file)
|
105 |
+
st.image(image, caption='Diagnosis', use_column_width=True)
|
106 |
+
model = timm.create_model(model_name='efficientnet_b0', pretrained=True,num_classes=4)
|
107 |
+
data_cfg = timm.data.resolve_data_config(model.pretrained_cfg)
|
108 |
+
transform = timm.data.create_transform(**data_cfg)
|
109 |
+
model_transforms = torchvision.transforms.Compose([transform])
|
110 |
+
transformed_image = model_transforms(image)
|
111 |
+
brain_model = torch.load('models/timm_skin_model.pth')
|
112 |
+
|
113 |
+
brain_model.eval()
|
114 |
+
with torch.inference_mode():
|
115 |
+
with st.progress(100):
|
116 |
+
|
117 |
+
#class_names = ['Glinomia','Meningomia','notumar','pituary']
|
118 |
+
prediction = torch.nn.functional.softmax(brain_model(transformed_image.unsqueeze(dim=0))[0], dim=0)
|
119 |
+
prediction_score, pred_label_idx = torch.topk(prediction, 1)
|
120 |
+
pred_label_idx.squeeze_()
|
121 |
+
predicted_label = idx_to_labels[str(pred_label_idx.item())]
|
122 |
+
st.write( f'Predicted Label: {predicted_label}')
|
123 |
+
if st.button('Know More'):
|
124 |
+
generator = pipeline("text-generation",model=text_model,tokenizer=tokenizer)
|
125 |
+
input_text = f"Patient has {predicted_label} and is advised to take the following medicines:"
|
126 |
+
with st.spinner('Generating Text'):
|
127 |
+
generator(input_text, max_length=300, do_sample=True, top_k=50, top_p=0.95, num_return_sequences=1)
|
128 |
+
st.markdown(generator(input_text, max_length=300, do_sample=True, top_k=50, top_p=0.95, num_return_sequences=1)[0]['generated_text'])
|
129 |
+
|
130 |
+
|
131 |
+
|
132 |
+
|
133 |
+
|
134 |
+
|
135 |
+
|
136 |
+
|
137 |
+
|
138 |
+
|
139 |
+
|
140 |
+
|
141 |
+
|
142 |
+
else:
|
143 |
+
st.success("Please upload an image file 🧠")
|
144 |
+
|
145 |
+
|
requirements.txt
ADDED
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
captum==0.6.0
|
2 |
+
lightning==2.0.1
|
3 |
+
matplotlib==3.6.3
|
4 |
+
Pillow==9.5.0
|
5 |
+
shap==0.41.0
|
6 |
+
streamlit==1.20.0
|
7 |
+
timm==0.6.13
|
8 |
+
torch==2.0.0
|
9 |
+
torchmetrics==0.11.4
|
10 |
+
torchvision==0.15.1
|
11 |
+
transformers==4.27.4
|
skin_labels.json
ADDED
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"0":"actinic keratoses and intraepithelial carcinoma",
|
3 |
+
"1": "basal cell carcinoma",
|
4 |
+
"2":"benign keratosis-like lesions",
|
5 |
+
"3": "dermatofibroma",
|
6 |
+
"4": "melanoma",
|
7 |
+
"5": "melanocytic nevi",
|
8 |
+
"6": "vascular lesions"
|
9 |
+
|
10 |
+
}
|