Updated application for corneal analysis with improved UI/UX and features

README.md

@@ -1,12 +1,54 @@
 ---
-title: Retina AI OCT
-emoji: 📉
-colorFrom: gray
-colorTo: gray
+title: Cornea AI Pentacam Analyzer
+emoji: 👁️
+colorFrom: blue
+colorTo: indigo
 sdk: streamlit
-sdk_version: 1.41.1
+sdk_version: 1.31.0
 app_file: app.py
 pinned: false
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+
+# Cornea AI Pentacam Analyzer
+
+Advanced corneal analysis and diagnostics tool developed by Dr. Verónica Gómez Calleja, cornea specialist.
+
+## Features
+
+- Comprehensive Pentacam scan analysis
+- Support for multiple file formats (PDF, PNG, JPG, JPEG)
+- Patient history and clinical data management
+- Comparative analysis between timepoints
+- Advanced corneal disease diagnostics including:
+  - Keratoconus assessment using ABCD staging
+  - Fuchs Endothelial Corneal Dystrophy (FECD) evaluation
+  - Subclinical corneal edema detection using Sun criteria
+
+## Setup
+
+1. Install Python 3.8 or higher
+2. Install dependencies:
+   ```bash
+   pip install -r requirements.txt
+   ```
+3. Create a `.env` file with your Google API key:
+   ```
+   GOOGLE_API_KEY=your_api_key_here
+   ```
+4. Run the application:
+   ```bash
+   streamlit run app.py
+   ```
+
+## Usage
+
+1. Enter patient information and clinical history
+2. Upload Pentacam scans (supported formats: PDF, PNG, JPG, JPEG)
+3. Choose between single timepoint analysis or comparison between timepoints
+4. Review the AI-powered analysis results
+
+## Note
+
+This tool is designed to assist medical professionals in their analysis and decision-making process. It should not replace professional medical judgment.

app.py

@@ -7,9 +7,9 @@ import PyPDF2
 import io
 from datetime import datetime
 
-# Page configuration must be the first Streamlit command
+# Page configuration
 st.set_page_config(
-    page_title="OCT Retina Analysis Assistant",
+    page_title="Cornea AI Pentacam Analyzer",
     page_icon="👁️",
     layout="wide",
     initial_sidebar_state="expanded"
@@ -30,147 +30,100 @@ st.markdown("""
     }
     .stButton>button {
         width: 100%;
-        background-color: #FF4B4B;
+        background-color: #2E86C1;
         color: white;
         padding: 0.5rem;
         margin-top: 1rem;
     }
     .credit-box {
         background-color: #f0f2f6;
-        padding: 1rem;
+        padding: 1.5rem;
         border-radius: 0.5rem;
         margin: 1rem 0;
+        border-left: 5px solid #2E86C1;
     }
     .header-box {
-        background-color: #FF4B4B;
+        background: linear-gradient(135deg, #2E86C1, #3498DB);
         padding: 2rem;
         border-radius: 0.5rem;
         color: white;
         margin-bottom: 2rem;
         text-align: center;
+        box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
     }
     .image-container {
         margin: 1rem 0;
         padding: 1rem;
         border-radius: 0.5rem;
-        background-color: #f0f2f6;
+        background-color: #f8f9fa;
+        border: 1px solid #e9ecef;
     }
     .analysis-container {
         margin-top: 1rem;
+        padding: 1.5rem;
+        border-radius: 0.5rem;
+        background-color: #f8f9fa;
+        border: 1px solid #e9ecef;
+    }
+    .patient-info {
+        background-color: #fff;
+        padding: 1.5rem;
+        border-radius: 0.5rem;
+        border: 1px solid #e9ecef;
+        margin-bottom: 1rem;
+    }
+    .upload-section {
+        background-color: #f8f9fa;
+        padding: 1.5rem;
+        border-radius: 0.5rem;
+        border: 1px dashed #2E86C1;
+        margin: 1rem 0;
+    }
+    .info-box {
+        background-color: #e1f5fe;
         padding: 1rem;
         border-radius: 0.5rem;
-        background-color: #f0f2f6;
+        margin: 0.5rem 0;
+        border-left: 3px solid #03a9f4;
     }
     </style>
 """, unsafe_allow_html=True)
 
 # System prompts
-SINGLE_TIMEPOINT_PROMPT = """You are an expert ophthalmologist specializing in interpreting macular Optical Coherence Tomography (OCT) scans. Your goal is to provide an accurate description and a possible diagnosis, supported by clear medical reasoning. These scans are from the same patient at a single timepoint. Please provide a comprehensive analysis.
-
-Step 1: Image Quality Assessment
-For each scan, describe the overall image quality, noting any artifacts or limitations that may affect your analysis.
-
-Step 2: Layer-by-Layer Analysis Across All Scans
-Analyze each of the retinal layers across all provided scans, describing:
-•	Thickness patterns: Note any variations or consistencies in layer thickness
-•	Morphological changes: Compare layer appearance across scans
-•	Reflectivity patterns: Identify any recurring patterns or changes
-•	Document abnormalities and their distribution across scans
-
-Step 3: Comprehensive Foveal Analysis
-Analyze the foveal region across all scans:
-•	Compare foveal contour and thickness
-•	Note any consistent or varying abnormalities
-•	Identify patterns of foveal involvement
-
-Step 4: Integrated Abnormality Assessment
-Provide a unified analysis of abnormalities across all scans:
-•	Distribution patterns
-•	Progression or variation in appearance
-•	Relationship between findings in different scans
-
-Step 5: Differential Diagnoses
-Based on the comprehensive analysis:
-•	List potential diagnoses supported by findings across multiple scans
-•	Explain how the pattern of findings supports each diagnosis
-•	Note any temporal or spatial progression that helps narrow the diagnosis
-
-Step 6: Most Likely Diagnosis
-Provide a unified diagnosis considering all scans:
-•	Explain how the combined findings support this diagnosis
-•	Discuss any progression or pattern that confirms the diagnosis
-•	Address any variations or inconsistencies
-
-Step 7: Recommendations
-Suggest:
-•	Additional tests or imaging if needed
-•	Follow-up scanning recommendations
-•	Treatment considerations based on the comprehensive analysis"""
-
-COMPARISON_PROMPT = """You are an expert ophthalmologist specializing in interpreting macular Optical Coherence Tomography (OCT) scans. Your goal is to provide an accurate description and a possible diagnosis, supported by clear medical reasoning. These scans are from the same patient at two different timepoints. Please provide a comprehensive analysis and comparison.
+CORNEA_ANALYSIS_PROMPT = """You are an expert ophthalmologist specializing in corneal diseases. A user will ask you questions about interpreting corneal imaging, specifically Pentacam, and about subclinical corneal edema.
 
-Step 1: Image Quality Assessment
-For each set of scans, describe the overall image quality, noting any artifacts or limitations that may affect your analysis.
+**Part 1: Pentacam Interpretation**
+• Analyze Pentacam imaging for diagnosing and monitoring corneal diseases
+• Assess key parameters: corneal shape, thickness, and elevation
+• Evaluate anterior and posterior corneal surfaces
+• Consider ABCD staging system and Cruces University Hospital progression score
 
-Step 2: Layer-by-Layer Comparison
-Compare the retinal layers between timepoints:
-•	Changes in thickness patterns
-•	Evolution of morphological features
-•	Alterations in reflectivity patterns
-•	Progression or regression of abnormalities
+**Part 2: Subclinical Corneal Edema Analysis**
+• Evaluate for Fuchs Endothelial Corneal Dystrophy (FECD)
+• Apply Sun criteria for subclinical corneal edema
+• Assess displacement of thinnest point
+• Check for loss of parallel isopachs
+• Evaluate focal depression of posterior surface
 
-Step 3: Foveal Evolution Analysis
-Compare the foveal region between timepoints:
-•	Changes in contour and thickness
-•	Evolution of abnormalities
-•	Progression or improvement patterns
+Please analyze the provided images and provide a detailed assessment."""
 
-Step 4: Disease Progression Assessment
-Analyze changes between timepoints:
-•	Quantify and describe changes in abnormalities
-•	Identify new or resolved findings
-•	Assess overall disease progression or improvement
+COMPARISON_PROMPT = """Compare the provided Pentacam scans from different timepoints, focusing on:
 
-Step 5: Treatment Response Evaluation
-If treatment was administered:
-•	Evaluate effectiveness
-•	Identify areas of improvement
-•	Note resistant or worsening areas
+1. Changes in corneal parameters:
+   • Thickness variations
+   • Topographic changes
+   • Elevation differences
+   • Progression indicators
 
-Step 6: Updated Diagnosis and Prognosis
-Based on the temporal comparison:
-•	Confirm or revise previous diagnosis
-•	Assess disease trajectory
-•	Provide prognostic insights
+2. Clinical significance:
+   • Disease progression/stability
+   • Treatment response
+   • Risk assessment
 
-Step 7: Recommendations
-Suggest:
-•	Treatment modifications if needed
-•	Follow-up interval
-•	Additional testing if required
-•	Preventive measures"""
-
-TREATMENT_GUIDELINES_PROMPT = """Based on the current diagnosis and findings, please provide evidence-based treatment recommendations following established ophthalmological guidelines. Consider:
-
-1. Standard of Care
-•	First-line treatments
-•	Alternative options
-•	Contraindications
-
-2. Treatment Plan
-•	Immediate interventions
-•	Long-term management
-•	Follow-up schedule
-
-3. Monitoring Parameters
-•	Key metrics to track
-•	Warning signs
-•	Success indicators
-
-4. Patient Education
-•	Lifestyle modifications
-•	Self-monitoring instructions
-•	Prevention strategies"""
+3. Recommendations:
+   • Follow-up interval
+   • Treatment modifications
+   • Additional testing needs"""
 
 def extract_pdf_text(pdf_file):
     """Extract text from uploaded PDF file"""
@@ -180,45 +133,39 @@ def extract_pdf_text(pdf_file):
         text += page.extract_text()
     return text
 
-def analyze_oct_images(images, timepoint=None, patient_data=None):
-    """Analyze OCT images with optional timepoint and patient data"""
+def analyze_cornea_images(images, timepoint=None, patient_data=None):
+    """Analyze corneal images with optional timepoint and patient data"""
     if timepoint:
-        prompt = f"{SINGLE_TIMEPOINT_PROMPT}\n\nTimepoint: {timepoint}\n"
+        prompt = f"{CORNEA_ANALYSIS_PROMPT}\n\nTimepoint: {timepoint}\n"
     else:
-        prompt = f"{SINGLE_TIMEPOINT_PROMPT}\n"
+        prompt = f"{CORNEA_ANALYSIS_PROMPT}\n"
     
     if patient_data:
         prompt += f"\nPatient Information:\n{patient_data}\n"
     
-    prompt += "\nPlease analyze these OCT scans:"
+    prompt += "\nPlease analyze these corneal scans:"
     content = [prompt] + images
     response = model.generate_content(content)
     return response.text
 
-def compare_oct_timepoints(images1, date1, images2, date2, patient_data=None):
-    """Compare OCT images from two timepoints"""
+def compare_timepoints(images1, date1, images2, date2, patient_data=None):
+    """Compare corneal images from two timepoints"""
     prompt = f"{COMPARISON_PROMPT}\n\nTimepoint 1: {date1}\nTimepoint 2: {date2}\n"
     
     if patient_data:
         prompt += f"\nPatient Information:\n{patient_data}\n"
     
-    prompt += "\nPlease compare these OCT scans:"
+    prompt += "\nPlease compare these corneal scans:"
     content = [prompt] + images1 + images2
     response = model.generate_content(content)
     return response.text
 
-def get_treatment_recommendations(diagnosis, findings):
-    """Get treatment recommendations based on guidelines"""
-    prompt = f"{TREATMENT_GUIDELINES_PROMPT}\n\nDiagnosis: {diagnosis}\nFindings: {findings}"
-    response = model.generate_content(prompt)
-    return response.text
-
 def main():
-    # Header with custom styling
+    # Header
     st.markdown("""
         <div class="header-box">
-            <h1>OCT Retina Analysis Assistant</h1>
-            
+            <h1>Cornea AI Pentacam Analyzer</h1>
+            <p style="font-size: 1.2em; margin-top: 1rem;">Advanced Corneal Analysis & Diagnostics</p>
         </div>
     """, unsafe_allow_html=True)
 
@@ -226,9 +173,10 @@ def main():
     st.markdown("""
         <div class="credit-box">
             <h3>About</h3>
-            <p>Developed by Dr. Fernando Ly</p>
-            <p>This tool assists in the analysis of OCT retina scans using advanced AI technology. 
-            It provides detailed layer analysis and potential diagnoses to support clinical decision-making.</p>
+            <p>Developed by Dr. Verónica Gómez Calleja</p>
+            <p>Cornea Specialist</p>
+            <p>This advanced tool assists in the analysis of Pentacam scans and corneal conditions using state-of-the-art AI technology. 
+            It provides comprehensive analysis of corneal parameters and supports clinical decision-making in keratoconus, FECD, and other corneal conditions.</p>
             <p><strong>Note:</strong> This tool is for assistance only and should not replace professional medical judgment.</p>
         </div>
     """, unsafe_allow_html=True)
@@ -237,153 +185,109 @@ def main():
     col1, col2 = st.columns([1, 1])
     
     with col1:
-        # Patient Data Section
+        # Patient Information Section
         st.markdown("### Patient Information")
-        patient_pdf = st.file_uploader("Upload Patient Data (PDF)", type=['pdf'])
-        patient_data = None
-        if patient_pdf:
-            patient_data = extract_pdf_text(patient_pdf)
-            with st.expander("View Patient Data"):
-                st.text(patient_data)
-        
-        # Scan Upload Section
-        st.markdown("### Upload OCT Scans")
-        timepoint_option = st.radio(
-            "Select scan type:",
-            ["Single Timepoint", "Two Timepoints for Comparison"]
-        )
+        with st.container():
+            st.markdown('<div class="patient-info">', unsafe_allow_html=True)
+            patient_name = st.text_input("Patient Name")
+            patient_id = st.text_input("Patient ID")
+            dob = st.date_input("Date of Birth")
+            diagnosis = st.text_input("Previous Diagnosis (if any)")
+            
+            # Clinical History
+            st.markdown("#### Clinical History")
+            history_text = st.text_area("Enter relevant clinical history")
+            
+            # Multiple PDF uploads for patient records
+            st.markdown("#### Additional Patient Records (PDF)")
+            patient_pdfs = st.file_uploader("Upload Patient Records", type=['pdf'], accept_multiple_files=True)
+            if patient_pdfs:
+                for pdf in patient_pdfs:
+                    with st.expander(f"View {pdf.name}"):
+                        st.text(extract_pdf_text(pdf))
+            st.markdown('</div>', unsafe_allow_html=True)
+
+        # Pentacam Scan Upload Section
+        st.markdown("### Upload Pentacam Scans")
+        st.markdown('<div class="upload-section">', unsafe_allow_html=True)
+        scan_type = st.radio("Select Analysis Type", ["Single Timepoint", "Compare Timepoints"])
         
-        if timepoint_option == "Single Timepoint":
-            uploaded_files = st.file_uploader(
-                "Choose OCT scans", 
-                type=['png', 'jpg', 'jpeg'],
-                accept_multiple_files=True,
-                key="single_timepoint"
-            )
+        if scan_type == "Single Timepoint":
+            scan_date = st.date_input("Scan Date")
+            uploaded_files = st.file_uploader("Upload Pentacam Scans", type=['png', 'jpg', 'jpeg', 'pdf'], accept_multiple_files=True)
             
             if uploaded_files:
-                scan_date = st.date_input("Scan Date")
-                st.markdown("### Uploaded Scans")
                 images = []
-                for idx, uploaded_file in enumerate(uploaded_files):
-                    with st.expander(f"OCT Scan {idx + 1}", expanded=True):
-                        image = Image.open(uploaded_file)
+                for file in uploaded_files:
+                    if file.type.startswith('image'):
+                        image = Image.open(file)
                         images.append(image)
-                        st.image(image, use_container_width=True, caption=f"OCT Scan {idx + 1}")
+                    elif file.type == 'application/pdf':
+                        # Handle PDF files here if needed
+                        st.warning(f"PDF processing for {file.name} will be implemented soon")
                 
-                if st.button("🔍 Analyze Scans"):
-                    with st.spinner("Analyzing OCT scans... Please wait."):
-                        try:
-                            analysis = analyze_oct_images(images, scan_date, patient_data)
-                            treatment_recs = get_treatment_recommendations(
-                                "Based on the analysis above",
-                                "See detailed findings above"
+                if images:
+                    st.markdown("#### Preview Uploaded Scans")
+                    cols = st.columns(len(images))
+                    for idx, (col, img) in enumerate(zip(cols, images)):
+                        with col:
+                            st.image(img, caption=f"Scan {idx + 1}", use_column_width=True)
+                    
+                    if st.button("Analyze Scans"):
+                        with st.spinner("Analyzing..."):
+                            analysis = analyze_cornea_images(
+                                images,
+                                timepoint=scan_date.strftime("%Y-%m-%d"),
+                                patient_data=f"Name: {patient_name}\nID: {patient_id}\nDOB: {str(dob)}\nDiagnosis: {diagnosis}\nHistory: {history_text}"
                             )
-                            
-                            with col2:
-                                st.markdown("### Analysis Results")
-                                st.markdown(f"""
-                                    <div class="analysis-container">
-                                        {analysis.replace(chr(10), '<br>')}
-                                    </div>
-                                """, unsafe_allow_html=True)
-                                
-                                st.markdown("### Treatment Recommendations")
-                                st.markdown(f"""
-                                    <div class="analysis-container">
-                                        {treatment_recs.replace(chr(10), '<br>')}
-                                    </div>
-                                """, unsafe_allow_html=True)
-                                
-                        except Exception as e:
-                            st.error(f"An error occurred during analysis: {str(e)}")
+                            st.markdown("### Analysis Results")
+                            st.markdown('<div class="analysis-container">', unsafe_allow_html=True)
+                            st.markdown(analysis)
+                            st.markdown('</div>', unsafe_allow_html=True)
         
-        else:  # Two Timepoints
-            st.markdown("#### First Timepoint")
-            files1 = st.file_uploader(
-                "Choose first set of OCT scans", 
-                type=['png', 'jpg', 'jpeg'],
-                accept_multiple_files=True,
-                key="timepoint1"
-            )
-            date1 = st.date_input("First Scan Date")
+        else:  # Compare Timepoints
+            col1, col2 = st.columns(2)
+            
+            with col1:
+                st.markdown("#### First Timepoint")
+                date1 = st.date_input("Date of First Scan")
+                files1 = st.file_uploader("Upload First Scans", type=['png', 'jpg', 'jpeg', 'pdf'], accept_multiple_files=True)
             
-            st.markdown("#### Second Timepoint")
-            files2 = st.file_uploader(
-                "Choose second set of OCT scans", 
-                type=['png', 'jpg', 'jpeg'],
-                accept_multiple_files=True,
-                key="timepoint2"
-            )
-            date2 = st.date_input("Second Scan Date")
+            with col2:
+                st.markdown("#### Second Timepoint")
+                date2 = st.date_input("Date of Second Scan")
+                files2 = st.file_uploader("Upload Second Scans", type=['png', 'jpg', 'jpeg', 'pdf'], accept_multiple_files=True)
             
             if files1 and files2:
-                st.markdown("### Uploaded Scans")
-                images1 = []
-                images2 = []
-                
-                st.markdown("#### First Timepoint Scans")
-                for idx, uploaded_file in enumerate(files1):
-                    with st.expander(f"OCT Scan {idx + 1} - First Timepoint", expanded=False):
-                        image = Image.open(uploaded_file)
-                        images1.append(image)
-                        st.image(image, use_container_width=True)
-                
-                st.markdown("#### Second Timepoint Scans")
-                for idx, uploaded_file in enumerate(files2):
-                    with st.expander(f"OCT Scan {idx + 1} - Second Timepoint", expanded=False):
-                        image = Image.open(uploaded_file)
-                        images2.append(image)
-                        st.image(image, use_container_width=True)
+                images1 = [Image.open(f) for f in files1 if f.type.startswith('image')]
+                images2 = [Image.open(f) for f in files2 if f.type.startswith('image')]
                 
-                if st.button("🔍 Compare Timepoints"):
-                    with st.spinner("Analyzing and comparing OCT scans... Please wait."):
-                        try:
-                            comparison = compare_oct_timepoints(
-                                images1, date1,
-                                images2, date2,
-                                patient_data
+                if images1 and images2:
+                    st.markdown("#### Preview First Timepoint")
+                    cols1 = st.columns(len(images1))
+                    for idx, (col, img) in enumerate(zip(cols1, images1)):
+                        with col:
+                            st.image(img, caption=f"Scan {idx + 1}", use_column_width=True)
+                    
+                    st.markdown("#### Preview Second Timepoint")
+                    cols2 = st.columns(len(images2))
+                    for idx, (col, img) in enumerate(zip(cols2, images2)):
+                        with col:
+                            st.image(img, caption=f"Scan {idx + 1}", use_column_width=True)
+                    
+                    if st.button("Compare Scans"):
+                        with st.spinner("Analyzing..."):
+                            comparison = compare_timepoints(
+                                images1, date1.strftime("%Y-%m-%d"),
+                                images2, date2.strftime("%Y-%m-%d"),
+                                patient_data=f"Name: {patient_name}\nID: {patient_id}\nDOB: {str(dob)}\nDiagnosis: {diagnosis}\nHistory: {history_text}"
                             )
-                            treatment_recs = get_treatment_recommendations(
-                                "Based on the comparison above",
-                                "See detailed findings above"
-                            )
-                            
-                            with col2:
-                                st.markdown("### Comparison Results")
-                                st.markdown(f"""
-                                    <div class="analysis-container">
-                                        {comparison.replace(chr(10), '<br>')}
-                                    </div>
-                                """, unsafe_allow_html=True)
-                                
-                                st.markdown("### Treatment Recommendations")
-                                st.markdown(f"""
-                                    <div class="analysis-container">
-                                        {treatment_recs.replace(chr(10), '<br>')}
-                                    </div>
-                                """, unsafe_allow_html=True)
-                                
-                        except Exception as e:
-                            st.error(f"An error occurred during analysis: {str(e)}")
-    
-    # Instructions in col2 if no files uploaded
-    if not (patient_pdf or (timepoint_option == "Single Timepoint" and uploaded_files) or 
-            (timepoint_option == "Two Timepoints for Comparison" and files1 and files2)):
-        with col2:
-            st.markdown("### Instructions")
-            st.markdown("""
-                1. Upload patient data PDF (optional)
-                2. Choose analysis type (single timepoint or comparison)
-                3. Upload OCT scans for selected timepoint(s)
-                4. Set scan dates
-                5. Click analyze/compare button
-                6. Review analysis and treatment recommendations
-                
-                Supported formats:
-                - Patient Data: PDF
-                - OCT Scans: PNG, JPG, JPEG
-            """)
+                            st.markdown("### Comparison Results")
+                            st.markdown('<div class="analysis-container">', unsafe_allow_html=True)
+                            st.markdown(comparison)
+                            st.markdown('</div>', unsafe_allow_html=True)
+        
+        st.markdown('</div>', unsafe_allow_html=True)
 
 if __name__ == "__main__":
     main() 

requirements.txt

@@ -1,5 +1,5 @@
-streamlit==1.32.0
-google-generativeai==0.3.2
-Pillow==10.2.0
-python-dotenv==1.0.1
-PyPDF2==3.0.1 
+streamlit>=1.31.0
+google-generativeai>=0.3.2
+Pillow>=10.2.0
+python-dotenv>=1.0.0
+PyPDF2>=3.0.0