Spaces:
Sleeping
Sleeping
Upload app.py
Browse files- src/app.py +444 -0
src/app.py
ADDED
|
@@ -0,0 +1,444 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# ============================================================
|
| 2 |
+
# DentaScribe β AI Dental Voice Assistant
|
| 3 |
+
# Role 6 β Demo & Presentation (Aparna)
|
| 4 |
+
# Streamlit front-end that runs the full pipeline end-to-end
|
| 5 |
+
# ============================================================
|
| 6 |
+
|
| 7 |
+
import streamlit as st
|
| 8 |
+
import tempfile
|
| 9 |
+
import os
|
| 10 |
+
import json
|
| 11 |
+
from datetime import datetime
|
| 12 |
+
|
| 13 |
+
# ββ Page config βββββββββββββββββββββββββββββββββββββββββββββ
|
| 14 |
+
st.set_page_config(
|
| 15 |
+
page_title="DentaScribe",
|
| 16 |
+
page_icon="π¦·",
|
| 17 |
+
layout="wide",
|
| 18 |
+
)
|
| 19 |
+
|
| 20 |
+
# ββ Custom CSS βββββββββββββββββββββββββββββββββββββββββββββββ
|
| 21 |
+
st.markdown("""
|
| 22 |
+
<style>
|
| 23 |
+
.main-header {
|
| 24 |
+
font-size: 2.8rem;
|
| 25 |
+
font-weight: 700;
|
| 26 |
+
color: #1a73e8;
|
| 27 |
+
text-align: center;
|
| 28 |
+
margin-bottom: 0.2rem;
|
| 29 |
+
}
|
| 30 |
+
.sub-header {
|
| 31 |
+
font-size: 1.1rem;
|
| 32 |
+
color: #555;
|
| 33 |
+
text-align: center;
|
| 34 |
+
margin-bottom: 2rem;
|
| 35 |
+
}
|
| 36 |
+
.step-box {
|
| 37 |
+
background: #f0f4ff;
|
| 38 |
+
border-left: 4px solid #1a73e8;
|
| 39 |
+
padding: 0.8rem 1.2rem;
|
| 40 |
+
border-radius: 6px;
|
| 41 |
+
margin-bottom: 0.6rem;
|
| 42 |
+
}
|
| 43 |
+
.entity-tag {
|
| 44 |
+
display: inline-block;
|
| 45 |
+
padding: 2px 10px;
|
| 46 |
+
border-radius: 12px;
|
| 47 |
+
font-size: 0.85rem;
|
| 48 |
+
font-weight: 600;
|
| 49 |
+
margin: 3px;
|
| 50 |
+
}
|
| 51 |
+
.tag-disease { background: #ffe0e0; color: #c0392b; }
|
| 52 |
+
.tag-chemical { background: #e0f0ff; color: #1565c0; }
|
| 53 |
+
.form-card {
|
| 54 |
+
background: #ffffff;
|
| 55 |
+
border: 1px solid #ddd;
|
| 56 |
+
border-radius: 10px;
|
| 57 |
+
padding: 1.5rem;
|
| 58 |
+
box-shadow: 0 2px 6px rgba(0,0,0,0.06);
|
| 59 |
+
}
|
| 60 |
+
.confidence-bar {
|
| 61 |
+
height: 10px;
|
| 62 |
+
border-radius: 5px;
|
| 63 |
+
background: linear-gradient(90deg, #1a73e8, #34a853);
|
| 64 |
+
}
|
| 65 |
+
</style>
|
| 66 |
+
""", unsafe_allow_html=True)
|
| 67 |
+
|
| 68 |
+
# ββ Header βββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 69 |
+
st.markdown('<div class="main-header">π¦· DentaScribe</div>', unsafe_allow_html=True)
|
| 70 |
+
st.markdown('<div class="sub-header">AI-powered voice assistant for dental clinics β speech β NER β auto-filled patient form</div>', unsafe_allow_html=True)
|
| 71 |
+
|
| 72 |
+
# ββ Pipeline loader (cached so models load once) βββββββββββββ
|
| 73 |
+
@st.cache_resource(show_spinner="Loading AI models β this takes ~30 s on first runβ¦")
|
| 74 |
+
def load_models():
|
| 75 |
+
"""
|
| 76 |
+
Load all three models into memory.
|
| 77 |
+
Role 1 β Andy : Whisper ASR
|
| 78 |
+
Role 2 β Ali : DistilBERT NER
|
| 79 |
+
Role 4 β Iva : DistilBERT Form Classifier
|
| 80 |
+
"""
|
| 81 |
+
from pipeline import load_whisper, load_ner, load_form_classifier
|
| 82 |
+
whisper_processor, whisper_model = load_whisper()
|
| 83 |
+
ner_pipe = load_ner()
|
| 84 |
+
form_tokenizer, form_model = load_form_classifier()
|
| 85 |
+
return whisper_processor, whisper_model, ner_pipe, form_tokenizer, form_model
|
| 86 |
+
|
| 87 |
+
|
| 88 |
+
# ββ Sidebar ββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 89 |
+
with st.sidebar:
|
| 90 |
+
st.image("https://img.icons8.com/color/96/tooth.png", width=80)
|
| 91 |
+
st.markdown("### DentaScribe")
|
| 92 |
+
st.markdown("**Deep Learning Group Project**")
|
| 93 |
+
st.markdown("---")
|
| 94 |
+
st.markdown("**Pipeline**")
|
| 95 |
+
st.markdown("""
|
| 96 |
+
1. ποΈ Audio Upload
|
| 97 |
+
2. π Whisper ASR *(Andy)*
|
| 98 |
+
3. π·οΈ NER *(Ali)*
|
| 99 |
+
4. ποΈ Data Engineering *(Varsha)*
|
| 100 |
+
5. π Form Classifier *(Iva)*
|
| 101 |
+
6. π Integration *(Koroush)*
|
| 102 |
+
7. π₯οΈ This Demo *(Aparna)*
|
| 103 |
+
""")
|
| 104 |
+
st.markdown("---")
|
| 105 |
+
st.markdown("**Patient ID**")
|
| 106 |
+
patient_id = st.text_input("Enter patient ID", value="P001", max_chars=20)
|
| 107 |
+
st.markdown("---")
|
| 108 |
+
st.markdown("**Input Mode**")
|
| 109 |
+
input_mode = st.radio(
|
| 110 |
+
"Choose how to provide input",
|
| 111 |
+
["ποΈ Record Live", "π Upload File", "π Demo Transcript"],
|
| 112 |
+
index=1,
|
| 113 |
+
label_visibility="collapsed",
|
| 114 |
+
)
|
| 115 |
+
demo_mode = input_mode == "π Demo Transcript"
|
| 116 |
+
|
| 117 |
+
DEMO_SCENARIOS = {
|
| 118 |
+
# ββ cleaning_form ββββββββββββββββββββββββββββββββββββββββ
|
| 119 |
+
"π¦ Gingivitis & Cleaning": (
|
| 120 |
+
"The patient presents with moderate gingivitis and early periodontitis. "
|
| 121 |
+
"We prescribed amoxicillin 500 mg three times daily for one week and "
|
| 122 |
+
"recommended ibuprofen for pain management. A thorough cleaning was performed."
|
| 123 |
+
),
|
| 124 |
+
# ββ root_canal_form ββββββββββββββββββββββββββββββββββββββ
|
| 125 |
+
"π¦· Root Canal": (
|
| 126 |
+
"Patient has severe pulpitis in the lower left molar with signs of apical abscess. "
|
| 127 |
+
"Root canal therapy was initiated today. We administered lidocaine for local anaesthesia "
|
| 128 |
+
"and prescribed metronidazole 400 mg twice daily for five days along with paracetamol "
|
| 129 |
+
"for post-procedure pain relief."
|
| 130 |
+
),
|
| 131 |
+
# ββ emergency_form βββββββββββββββββββββββββββββββββββββββ
|
| 132 |
+
"π¨ Emergency Visit": (
|
| 133 |
+
"Emergency visit for acute pericoronitis around the lower wisdom tooth. "
|
| 134 |
+
"Significant swelling and trismus observed. Patient was given a chlorhexidine "
|
| 135 |
+
"mouthwash prescription and amoxicillin-clavulanate 875 mg twice daily. "
|
| 136 |
+
"Extraction of the third molar is recommended next week."
|
| 137 |
+
),
|
| 138 |
+
# ββ extraction_form ββββββββββββββββββββββββββββββββββββββ
|
| 139 |
+
"π§ Tooth Extraction": (
|
| 140 |
+
"Extraction of the upper left second molar performed today due to severe dental caries "
|
| 141 |
+
"and irreversible pulpitis. Local anaesthesia with articaine was administered. "
|
| 142 |
+
"Patient prescribed amoxicillin 500 mg three times daily and ibuprofen 400 mg "
|
| 143 |
+
"for swelling. Post-extraction instructions provided."
|
| 144 |
+
),
|
| 145 |
+
# ββ filling_form βββββββββββββββββββββββββββββββββββββββββ
|
| 146 |
+
"πͺ₯ Cavity Filling": (
|
| 147 |
+
"Composite resin filling placed on the lower right first molar. "
|
| 148 |
+
"Patient presented with moderate dental caries extending to the dentine. "
|
| 149 |
+
"No signs of pulpitis detected. Fluoride varnish applied after the procedure. "
|
| 150 |
+
"Patient advised to avoid hard foods for 24 hours."
|
| 151 |
+
),
|
| 152 |
+
# ββ crown_bridge_form ββββββββββββββββββββββββββββββββββββ
|
| 153 |
+
"π Crown & Bridge": (
|
| 154 |
+
"The patient requires a porcelain crown on the upper right premolar following "
|
| 155 |
+
"a cracked tooth diagnosis. Impressions were taken today. Temporary crown placed "
|
| 156 |
+
"and patient prescribed ibuprofen 400 mg as needed for sensitivity. "
|
| 157 |
+
"Final crown fitting is scheduled in two weeks."
|
| 158 |
+
),
|
| 159 |
+
# ββ periodontal_form βββββββββββββββββββββββββββββββββββββ
|
| 160 |
+
"πΏ Periodontal Treatment": (
|
| 161 |
+
"Deep periodontal scaling performed on all four quadrants. "
|
| 162 |
+
"Patient has chronic periodontitis with 5 to 7 mm pocket depths. "
|
| 163 |
+
"Doxycycline 100 mg once daily prescribed for two weeks. "
|
| 164 |
+
"Chlorhexidine gel applied subgingivally. Follow-up in six weeks."
|
| 165 |
+
),
|
| 166 |
+
# ββ orthodontic_form βββββββββββββββββββββββββββββββββββββ
|
| 167 |
+
"π Orthodontic Consultation": (
|
| 168 |
+
"Initial orthodontic consultation for Class II malocclusion with moderate crowding. "
|
| 169 |
+
"Patient presents with dental fluorosis on the upper incisors. "
|
| 170 |
+
"Treatment plan includes fixed braces over 18 months. "
|
| 171 |
+
"Fluoride gel application recommended before bonding appointment."
|
| 172 |
+
),
|
| 173 |
+
# ββ whitening_form βββββββββββββββββββββββββββββββββββββββ
|
| 174 |
+
"β¨ Teeth Whitening": (
|
| 175 |
+
"Patient requested professional teeth whitening for mild to moderate tooth discolouration. "
|
| 176 |
+
"In-office bleaching performed using hydrogen peroxide gel. "
|
| 177 |
+
"Mild dentinal hypersensitivity noted post-procedure. "
|
| 178 |
+
"Prescribed fluoride toothpaste and advised to avoid coffee and tea for 48 hours."
|
| 179 |
+
),
|
| 180 |
+
# ββ consultation_form ββββββββββββββββββββββββββββββββββββ
|
| 181 |
+
"π General Consultation": (
|
| 182 |
+
"New patient consultation. Patient reports intermittent toothache and bleeding gums. "
|
| 183 |
+
"Examination reveals mild gingivitis and early signs of bruxism. "
|
| 184 |
+
"Dental radiographs taken. Recommended night guard for bruxism and "
|
| 185 |
+
"chlorhexidine mouthwash twice daily. Full treatment plan to be discussed at next visit."
|
| 186 |
+
),
|
| 187 |
+
}
|
| 188 |
+
|
| 189 |
+
if "custom_transcript" not in st.session_state:
|
| 190 |
+
st.session_state["custom_transcript"] = list(DEMO_SCENARIOS.values())[0]
|
| 191 |
+
|
| 192 |
+
if demo_mode:
|
| 193 |
+
preset = st.selectbox(
|
| 194 |
+
"π Load a demo scenario",
|
| 195 |
+
options=["β type your own β"] + list(DEMO_SCENARIOS.keys()),
|
| 196 |
+
)
|
| 197 |
+
if preset != "β type your own β":
|
| 198 |
+
st.session_state["custom_transcript"] = DEMO_SCENARIOS[preset]
|
| 199 |
+
|
| 200 |
+
SAMPLE_TRANSCRIPT = st.text_area(
|
| 201 |
+
"Transcript (edit freely)",
|
| 202 |
+
key="custom_transcript",
|
| 203 |
+
height=160,
|
| 204 |
+
)
|
| 205 |
+
|
| 206 |
+
# ββ Main content βββββββββββββββββββββββββββββββββββββββββββββ
|
| 207 |
+
col_left, col_right = st.columns([1, 1], gap="large")
|
| 208 |
+
|
| 209 |
+
with col_left:
|
| 210 |
+
audio_file = None
|
| 211 |
+
recorded_audio = None
|
| 212 |
+
|
| 213 |
+
if input_mode == "ποΈ Record Live":
|
| 214 |
+
st.markdown("### Step 1 β Record Live Audio")
|
| 215 |
+
recorded_audio = st.audio_input("ποΈ Press to record the consultation")
|
| 216 |
+
if recorded_audio is not None:
|
| 217 |
+
st.success("β
Recording captured β press play to review, then run the pipeline.")
|
| 218 |
+
elif input_mode == "π Upload File":
|
| 219 |
+
st.markdown("### Step 1 β Upload Audio")
|
| 220 |
+
audio_file = st.file_uploader(
|
| 221 |
+
"Upload a WAV, MP3 or M4A recording of the dental consultation",
|
| 222 |
+
type=["wav", "mp3", "m4a"],
|
| 223 |
+
)
|
| 224 |
+
if audio_file is not None:
|
| 225 |
+
st.markdown("**π Preview β play to listen along while the model transcribes:**")
|
| 226 |
+
st.audio(audio_file)
|
| 227 |
+
else: # Demo Transcript
|
| 228 |
+
st.markdown("### Step 1 β Demo Transcript")
|
| 229 |
+
st.info("π Demo mode: using built-in sample transcript β no audio needed.")
|
| 230 |
+
|
| 231 |
+
run_btn = st.button("βΆ Run DentaScribe Pipeline", type="primary", use_container_width=True)
|
| 232 |
+
|
| 233 |
+
# ββ Pipeline execution βββββββββββββββββββββββββββββββββββββββ
|
| 234 |
+
if run_btn:
|
| 235 |
+
# Pick the active audio source based on input mode
|
| 236 |
+
audio_source = recorded_audio if input_mode == "ποΈ Record Live" else audio_file
|
| 237 |
+
if input_mode != "π Demo Transcript" and audio_source is None:
|
| 238 |
+
if input_mode == "ποΈ Record Live":
|
| 239 |
+
st.error("Please record audio first, or switch to Upload File / Demo Transcript.")
|
| 240 |
+
else:
|
| 241 |
+
st.error("Please upload an audio file or switch input mode.")
|
| 242 |
+
st.stop()
|
| 243 |
+
|
| 244 |
+
# Read bytes once so we can both transcribe and replay them in Results
|
| 245 |
+
audio_bytes = audio_source.getvalue() if audio_source is not None else None
|
| 246 |
+
|
| 247 |
+
# Load models
|
| 248 |
+
try:
|
| 249 |
+
whisper_processor, whisper_model, ner_pipe, form_tokenizer, form_model = load_models()
|
| 250 |
+
except Exception as e:
|
| 251 |
+
st.error(f"Model loading failed: {e}")
|
| 252 |
+
st.stop()
|
| 253 |
+
|
| 254 |
+
from pipeline import (
|
| 255 |
+
transcribe_audio, extract_entities,
|
| 256 |
+
classify_and_fill_form, save_to_filing_system,
|
| 257 |
+
)
|
| 258 |
+
|
| 259 |
+
# ββ Step 1: Transcription ββββββββββββββββββββββββββββββββ
|
| 260 |
+
with st.spinner("ποΈ Transcribing audioβ¦"):
|
| 261 |
+
if input_mode == "π Demo Transcript":
|
| 262 |
+
transcript = SAMPLE_TRANSCRIPT
|
| 263 |
+
st.success("β
Demo transcript loaded.")
|
| 264 |
+
else:
|
| 265 |
+
if input_mode == "ποΈ Record Live":
|
| 266 |
+
suffix = ".wav"
|
| 267 |
+
else:
|
| 268 |
+
suffix = "." + audio_source.name.split(".")[-1]
|
| 269 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as tmp:
|
| 270 |
+
tmp.write(audio_bytes)
|
| 271 |
+
tmp_path = tmp.name
|
| 272 |
+
try:
|
| 273 |
+
transcript = transcribe_audio(tmp_path, whisper_processor, whisper_model)
|
| 274 |
+
st.success("β
Transcription complete.")
|
| 275 |
+
finally:
|
| 276 |
+
os.unlink(tmp_path)
|
| 277 |
+
|
| 278 |
+
# ββ Step 2: NER ββββββββββββββββββββββββββββββββββββββββββ
|
| 279 |
+
with st.spinner("π·οΈ Running Named Entity Recognitionβ¦"):
|
| 280 |
+
entities = extract_entities(transcript, ner_pipe)
|
| 281 |
+
st.success(f"β
{len(entities)} entities detected.")
|
| 282 |
+
|
| 283 |
+
# ββ Step 3: Form Classification + Autofill βββββββββββββββ
|
| 284 |
+
with st.spinner("π Classifying form & auto-filling fieldsβ¦"):
|
| 285 |
+
form = classify_and_fill_form(transcript, entities, form_tokenizer, form_model)
|
| 286 |
+
st.success("β
Form classified and filled.")
|
| 287 |
+
|
| 288 |
+
# ββ Step 4: Filing βββββββββββββββββββββββββββββββββββββββ
|
| 289 |
+
with st.spinner("ποΈ Saving to filing systemβ¦"):
|
| 290 |
+
saved_path = save_to_filing_system(patient_id, form)
|
| 291 |
+
st.success(f"β
Saved β `{saved_path}`")
|
| 292 |
+
|
| 293 |
+
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 294 |
+
# RESULTS
|
| 295 |
+
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 296 |
+
st.markdown("---")
|
| 297 |
+
st.markdown("## Results")
|
| 298 |
+
|
| 299 |
+
r1, r2, r3 = st.columns(3)
|
| 300 |
+
r1.metric("Form Type", form.get("form_type", "β").replace("_", " ").title())
|
| 301 |
+
r2.metric("Confidence", f"{form.get('confidence', 0)*100:.1f}%")
|
| 302 |
+
r3.metric("Entities Found", len(entities))
|
| 303 |
+
|
| 304 |
+
st.markdown("---")
|
| 305 |
+
|
| 306 |
+
tab1, tab2, tab3 = st.tabs(["π Transcript & Entities", "π Auto-filled Form", "ποΈ Raw JSON"])
|
| 307 |
+
|
| 308 |
+
# Tab 1 β Transcript & NER
|
| 309 |
+
with tab1:
|
| 310 |
+
if audio_bytes is not None:
|
| 311 |
+
st.markdown("#### π Original Audio")
|
| 312 |
+
st.audio(audio_bytes)
|
| 313 |
+
st.caption("Play back the original recording while reading the transcript to spot ASR errors.")
|
| 314 |
+
st.markdown("#### Transcript")
|
| 315 |
+
st.info(transcript)
|
| 316 |
+
|
| 317 |
+
st.markdown("#### Detected Entities")
|
| 318 |
+
if entities:
|
| 319 |
+
html_tags = ""
|
| 320 |
+
for ent in entities:
|
| 321 |
+
label = ent.get("entity_group", ent.get("label", ""))
|
| 322 |
+
word = ent.get("word", "")
|
| 323 |
+
score = ent.get("score", 0)
|
| 324 |
+
if "Disease" in label or "DISEASE" in label:
|
| 325 |
+
css = "tag-disease"
|
| 326 |
+
icon = "π¦ "
|
| 327 |
+
elif "Chemical" in label or "CHEM" in label:
|
| 328 |
+
css = "tag-chemical"
|
| 329 |
+
icon = "π"
|
| 330 |
+
else:
|
| 331 |
+
css = "tag-chemical"
|
| 332 |
+
icon = "π·οΈ"
|
| 333 |
+
html_tags += (
|
| 334 |
+
f'<span class="entity-tag {css}">'
|
| 335 |
+
f'{icon} {word} <small>({label} {score:.0%})</small>'
|
| 336 |
+
f'</span>'
|
| 337 |
+
)
|
| 338 |
+
st.markdown(html_tags, unsafe_allow_html=True)
|
| 339 |
+
|
| 340 |
+
# Entity table
|
| 341 |
+
st.markdown("#### Entity Details")
|
| 342 |
+
import pandas as pd
|
| 343 |
+
df = pd.DataFrame([
|
| 344 |
+
{
|
| 345 |
+
"Word": e.get("word", ""),
|
| 346 |
+
"Label": e.get("entity_group", e.get("label", "")),
|
| 347 |
+
"Confidence": f"{e.get('score', 0):.2%}",
|
| 348 |
+
}
|
| 349 |
+
for e in entities
|
| 350 |
+
])
|
| 351 |
+
st.dataframe(df, use_container_width=True, hide_index=True)
|
| 352 |
+
else:
|
| 353 |
+
st.warning("No entities detected in transcript.")
|
| 354 |
+
|
| 355 |
+
# Tab 2 β Auto-filled form
|
| 356 |
+
with tab2:
|
| 357 |
+
st.markdown("#### Patient Form")
|
| 358 |
+
st.markdown('<div class="form-card">', unsafe_allow_html=True)
|
| 359 |
+
|
| 360 |
+
fc1, fc2 = st.columns(2)
|
| 361 |
+
with fc1:
|
| 362 |
+
st.markdown(f"**Patient ID:** `{patient_id}`")
|
| 363 |
+
st.markdown(f"**Form Type:** {form.get('form_type','β').replace('_',' ').title()}")
|
| 364 |
+
st.markdown(f"**Date / Time:** {form.get('timestamp', datetime.now().strftime('%Y-%m-%d %H:%M'))}")
|
| 365 |
+
st.markdown(f"**Status:** {form.get('status','β').replace('_',' ').title()}")
|
| 366 |
+
with fc2:
|
| 367 |
+
conf = form.get("confidence", 0)
|
| 368 |
+
st.markdown(f"**Classifier Confidence:** {conf*100:.1f}%")
|
| 369 |
+
st.markdown(
|
| 370 |
+
f'<div class="confidence-bar" style="width:{conf*100:.0f}%;"></div>',
|
| 371 |
+
unsafe_allow_html=True,
|
| 372 |
+
)
|
| 373 |
+
|
| 374 |
+
st.markdown("---")
|
| 375 |
+
|
| 376 |
+
fc3, fc4 = st.columns(2)
|
| 377 |
+
with fc3:
|
| 378 |
+
st.markdown("**Diagnosis / Condition**")
|
| 379 |
+
diag = form.get("diagnosis", [])
|
| 380 |
+
if diag:
|
| 381 |
+
for d in (diag if isinstance(diag, list) else [diag]):
|
| 382 |
+
st.markdown(f"- π¦ {d}")
|
| 383 |
+
else:
|
| 384 |
+
st.markdown("_Not detected_")
|
| 385 |
+
|
| 386 |
+
with fc4:
|
| 387 |
+
st.markdown("**Medication / Treatment**")
|
| 388 |
+
meds = form.get("medication", [])
|
| 389 |
+
if meds:
|
| 390 |
+
for m in (meds if isinstance(meds, list) else [meds]):
|
| 391 |
+
st.markdown(f"- π {m}")
|
| 392 |
+
else:
|
| 393 |
+
st.markdown("_Not detected_")
|
| 394 |
+
|
| 395 |
+
st.markdown("---")
|
| 396 |
+
st.markdown("**Full Transcript (recorded)**")
|
| 397 |
+
st.text_area("", value=form.get("original_text", transcript), height=120, disabled=True)
|
| 398 |
+
|
| 399 |
+
st.markdown('</div>', unsafe_allow_html=True)
|
| 400 |
+
|
| 401 |
+
# Tab 3 β Raw JSON
|
| 402 |
+
with tab3:
|
| 403 |
+
st.markdown("#### Raw Form JSON (saved to filing system)")
|
| 404 |
+
st.json(form)
|
| 405 |
+
st.markdown("#### All Detected Entities (JSON)")
|
| 406 |
+
st.json(entities)
|
| 407 |
+
|
| 408 |
+
else:
|
| 409 |
+
# Landing state β show pipeline diagram
|
| 410 |
+
with col_right:
|
| 411 |
+
st.markdown("### How it works")
|
| 412 |
+
steps = [
|
| 413 |
+
("ποΈ", "Step 1 β Speech Recognition", "Whisper (fine-tuned) converts dental audio to text *(Andy)*"),
|
| 414 |
+
("π·οΈ", "Step 2 β Named Entity Recognition", "DistilBERT NER extracts diseases & medications *(Ali)*"),
|
| 415 |
+
("π", "Step 3 β Form Classification", "DistilBERT classifier picks the right form type *(Iva)*"),
|
| 416 |
+
("ποΈ", "Step 4 β Filing System", "JSON form saved by patient ID + date *(Koroush)*"),
|
| 417 |
+
]
|
| 418 |
+
for icon, title, desc in steps:
|
| 419 |
+
st.markdown(
|
| 420 |
+
f'<div class="step-box">'
|
| 421 |
+
f'<strong>{icon} {title}</strong><br><span style="color:#555">{desc}</span>'
|
| 422 |
+
f'</div>',
|
| 423 |
+
unsafe_allow_html=True,
|
| 424 |
+
)
|
| 425 |
+
|
| 426 |
+
st.markdown("---")
|
| 427 |
+
st.markdown("### Quick Start")
|
| 428 |
+
st.markdown("""
|
| 429 |
+
1. Enter a **Patient ID** in the sidebar (e.g. `P001`)
|
| 430 |
+
2. Upload a **WAV/MP3** recording of a dental consultation
|
| 431 |
+
β or tick **Demo Mode** to use the built-in sample
|
| 432 |
+
3. Click **βΆ Run DentaScribe Pipeline**
|
| 433 |
+
4. See the transcript, detected entities, and the auto-filled patient form
|
| 434 |
+
""")
|
| 435 |
+
|
| 436 |
+
# ββ Footer βββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 437 |
+
st.markdown("---")
|
| 438 |
+
st.markdown(
|
| 439 |
+
"<div style='text-align:center;color:#999;font-size:0.8rem;'>"
|
| 440 |
+
"DentaScribe Β· Deep Learning Group Project Β· "
|
| 441 |
+
"Andy Β· Ali Β· Varsha Β· Iva Β· Koroush Β· Aparna"
|
| 442 |
+
"</div>",
|
| 443 |
+
unsafe_allow_html=True,
|
| 444 |
+
)
|