Initial

README.md

@@ -0,0 +1 @@
+A note extraction app hosted on Hugging Face Spaces.

prompts.py

@@ -0,0 +1,505 @@
+info_prompt = """For each clinical note, extract the following fields into a structured XML format.
+For each of the following fields, return the value if it exists in the notes, otherwise do not return anything between tags.
+For example, if the notes mention that the patient's name is "John Doe", then the output should be <patient_name>John Doe</patient_name>.
+Otherwise, if the notes do not mention the patient's name, then do not return anything between <patient_name> and </patient_name>.
+Additionally, add a <reasoning> tag with the reasoning for why you chose the value you did. This reasoning should be specific to the note and the patient, and if a field is found, it should contain a brief verbatim quote from the notes that is used to justify the value.
+
+- patient_name
+    - The patient's full name (first and last name).
+    - For example:
+        <patient_name>
+            <reasoning>[REASONING]</reasoning>
+            <first_name>John</first_name>
+            <last_name>Doe</last_name>
+        </patient_name>
+- date_of_birth
+    - The patient's date of birth in the format YYYY-MM-DD.
+    - For example:
+        <date_of_birth>
+            <reasoning>[REASONING]</reasoning>
+            <date>1990-01-01</date>
+        </date_of_birth>   
+- sex
+    - The patient's sex (M or F).
+    - For example:
+        <sex>
+            <reasoning>[REASONING]</reasoning>
+            <sex>M</sex>
+        </sex>
+- traditional_chemo
+    - Any traditional chemotherapy drugs the patient has taken or has been prescribed.
+    - Within the tags, put a list of all the traditional chemotherapy drugs the patient has taken or has been prescribed as well as the date, if specified.
+    - For example:
+        <traditional_chemo_any_time>
+            <reasoning>[REASONING]</reasoning>
+            <drug>Doxorubicin/Adriamycin</drug>
+            <date>2021-01-01</date>
+        </traditional_chemo_any_time>
+    - The following are the traditional chemotherapy drugs you should look for:
+        - Doxorubicin/Adriamycin
+        - Carboplatin
+        - Vinblastine
+        - Chlorambucil/Leukeran
+        - Lomustine/CCNU
+        - Mitoxantrone
+        - Cyclophosphamide/Cytoxan
+        - Vinorelbine
+        - Vincristine
+        - CHOP protocol
+        - Actinomycin
+        - VAC Protocol
+        - Tanovea
+        - L-asparaginase
+        - Melphalan
+        - Satraplatin
+        - Epirubicin
+        - Neoplasene
+        - MOPP Chemotherapy
+        - Satraplatin metronomic
+        - Gemcitabine/Gemzar
+        - Fluorouracil (5-FU)
+        - Laverdia
+        - Temodar
+        - Unspecified traditional chemo
+        - No Traditional Chemo to-date
+        - No traditional chemo reported - validated
+        - Tamozolamide/Temodar
+        - Unknown
+        - Cisplatin
+        - Mustargen
+        - Procarbazine
+        - Mitotane
+
+- other_cancer_treatments
+    - Any other cancer treatments the patient has taken or has been prescribed, according to the list provided below.
+    - Within the tags, put a list of all the other cancer treatments the patient has taken or has been prescribed as well as the date, if specified.
+    - For example:
+        <other_cancer_treatments>
+            <reasoning>[REASONING]</reasoning>
+            <treatment>Radiation Therapy</treatment>
+            <date>2021-01-01</date>
+        </other_cancer_treatments>
+    - The following are the other cancer treatments you should look for:
+        - Radiation Therapy
+        - Palladia/Toceranib
+        - Melanoma Vaccine (Oncept)
+        - Electrochemotherapy
+        - Masatinib
+        - Autologous Vaccine/Torigen/Ardent
+        - Lapatinib
+        - I'm Yunity
+        - Yunnan Baiyao
+        - Previcox
+        - Yale Vaccine
+        - Rapamycin
+        - Listeria Vaccine
+        - Imatinib
+        - Trametinib
+        - Zoledronate
+        - Dexrazoxane/Zinecard
+        - Firocoxib
+        - Olaparib
+        - Dasatinib
+        - Vorinostat
+        - Mistletoe Therapy
+        - EGFR Vaccine
+        - No other cancer treatments reported - validated
+        - Palbociclib
+        - No other cancer treatments reported to-date
+        - Sorafenib
+        - Nanoparticle Infusion
+        - Nanoparticle Laser
+        - Laser Therapy
+        - Stelfonta
+        - Tanovea
+        - T-Cell infusions
+        - Losartan
+        - Naltrexone
+        - Immunoregulin
+        - Papilloma Vaccine
+        - Gilvetmab
+        - Unknown
+
+- other_conmeds
+    - Any other concomitant medications the patient has taken or has been prescribed.
+    - Within the tags, put a list of all the other concomitant medications the patient has taken or has been prescribed as well as the date, if specified.
+    - For example:
+        <other_conmeds>
+            <reasoning>[REASONING]</reasoning>
+            <medication>Aspirin</medication>
+            <date>2021-01-01</date>
+        </other_conmeds>
+    - The following are the other concomitant medications you should look for:
+        - Piroxicam
+        - Gabapentin
+        - Carprofen/Rimadyl
+        - Denamarin
+        - Ursodiol
+        - Clavamox
+        - Cerenia/Maropitant
+        - Ondansetron/Zofran
+        - Meloxicam/Metacam
+        - Pimobendan/Vetmedin
+        - Losartan/Cozaar
+        - Capromorelin/Entyce
+        - Cetirizine/Zyrtec
+        - Tacrolimus
+        - Codeine
+        - Telmisartan
+        - Buprenorphine
+        - Apoquel/Oclacitinib
+        - Imuquin
+        - Amlodipine
+        - Loratadine/Claritin
+        - Benazepril
+        - Metronidazole/Flagyl
+        - Prednisone
+        - Adequan
+        - Convenia
+        - B12 Injections
+        - Cisapride
+        - Budesonide
+        - Hepatoclear
+        - Dasaquin
+        - Cytopoint Injections
+        - Glucosamine
+        - Famotidine/Pepcid
+        - Fish Oil
+        - Omeprazole/Prilosec
+        - Mirtazapine
+        - Meclizine
+        - Amantadine
+        - Cortisone
+        - Pentoxifylline
+        - Ligaplex
+        - Reishi Mushroom
+        - Immune Builder
+        - CAS Multimushroom
+        - Ketamine Injections
+        - Vitamin E
+        - Trazadone
+        - Phenobarbitol
+        - Tylan Powder
+        - Temaril-P
+        - Acepromazine
+        - Sulfasalazine
+        - Keppra
+        - Turkey Tail Mushroom
+        - Furosemide/Lasix
+        - Tramadol
+        - Ciprofloxacin
+        - Trilostane
+        - Naturvet Vitapet Vitamins
+        - Glycoflex
+        - Entederm
+        - Aluminum Hydroxide
+        - Deramaxx
+        - Doxycycline
+        - Sulcrafate
+        - Diphenhydramine
+        - Fluoxetine
+        - Nexgard
+        - Reglan/Metoclopramide
+        - Thyroxine
+        - Clindamycin
+        - Cephalexin
+        - Enalapril
+        - CBD
+        - Denosyl
+        - Galliprant
+        - Methadone
+        - Cobalequin
+        - Azodyl
+        - FortiFlora
+        - Propectalin Paste
+        - Dexamethasone
+        - Ampicillin
+        - Coriolus mushroom
+        - Oxycodone
+        - Cyproheptadine
+        - Sotalol
+        - Enrofloxacin/Baytril
+        - Amiikacin
+        - Misoprostol
+        - Chlorhexadine
+        - Neomycin
+        - Visbiome
+        - Tranexamic Acid
+        - Proin
+        - Tobramycin
+        - Avmaquin
+        - Cosyntropin (Cortosyn)
+        - Vetoryl
+        - Metoclopromide
+        - Phenylpropanolamine HCl
+        - Cosequin
+        - Osteoflex
+        - Hepato TruBenefits
+        - Rx Clay
+        - Metamucil Powder
+        - Osteo-Tru Benefits
+        - Oat Glycerite
+        - Cholodin
+        - Proviable
+        - Supplements
+        - Wuffles Joint Supplement
+        - Firocoxib/Previcox
+        - Tylosin
+        - Barium Suspension
+        - Optimmune Ointment
+        - NeoPolyDex Solution
+        - Endosorb
+        - Augmentin
+        - Butorphanol/Dolorex
+        - Prazosin
+        - Traumeel/T-Relief
+        - Deracoxib
+        - Triamcinolone
+        - Probiotic
+        - Hydrocodone
+        - Lactulose
+        - Methocarbamol
+        - Cranberry Pills
+        - Eye Meds
+        - Levothyroxine
+        - Calcitriol
+        - TMS trimethoprim sulfamethoxazole
+        - Allergy Antigen Injections
+        - Propranolol
+        - Flexadin
+        - Interceptor
+        - Thorn SAT
+        - Megaflora
+        - Pregabalin
+        - Canalevia
+        - Cefpodoxime
+        - Melatonin
+        - Phenylephrine
+        - Amoxicillin
+        - Arnica/T-Relief
+        - Aminocaproic Acid
+        - Fluconazole
+        - Gastrafate
+        - Silver Sulfadiazine
+        - Mupirocin
+        - Marbofloxacin/Zeniquin
+        - Psyllium Husk
+        - Chlorpheniramine
+        - Tagamet
+        - Multi-vitamin
+        - D-mannose/cranberry
+        - Darbepoetin
+        - Soloxine
+        - Thuja Occidentalis
+        - Pantoprazole
+        - Normosol-R
+        - Nitrofurantoin
+        - Sildenafil
+        - Hydromorphone
+        - Terbinafine
+        - Sucralfate
+        - Clopidogrel
+        - EndoBlend
+        - Omega Benefts
+        - Dexmedetomidine
+        - Levetiracetam
+        - Diethylstilbesterol
+        - Nattokinase
+        - D3 supplement
+        - Modified Chai Hu Jia Long Gu Mu Li Tang supplement
+        - Power mushrooms
+        - super greens supplement
+        - Sertraline/Zoloft
+        - Mushroom Supplement
+        - Simparica Trio/Sarolaner, moxidectin, and pyrantel
+        - 5DMM
+        - Joint Supplements
+        - Vetericyn
+        - Milk Thistle
+        - S-Adenosyl methionine
+        - Cimetidine
+        - Silver Entro Dex
+        - Desmopressin
+        - Alpha lipoic acid
+        - Unasyn
+        - Panacur/Fenbendazole
+        - Xiao Chai Hu Tang
+        - Incurin
+        - Dextrose
+        - Fresh Frozen Plasma
+        - Pamidronate Infusion
+        - Curcumin
+        - Diazoxide
+        - Clavacillin
+        - Tetracycline
+        - B9 Folic Acid
+        - Prednisolone
+        - Cyclosporine
+        - Ketaconazole
+        - Novolin-N
+        - Zonisamide
+        - Gentamicin/Phenylephrine nasal drops
+        - Stool Softener
+        - Amitriptyline
+        - Moxifloxacin
+        - Gemfibrozil
+        - Taurine
+        - Mometamax
+        - Heartgard
+        - Green Lipped Mussel Powder
+        - Chlorella Powder
+        - BioSponge
+        - Folate
+        - Cobalamine
+        - Diazepam
+        - GenOne
+        - Phenoxybenzamine
+        - Flumethrin and Imidacloprid/Seresto
+        - Forte Ion Gut Health
+        - Dispel Stasis
+        - Blood Remaker + Immune Support with Mushrooms
+        - Pet Tab
+        - Omega 3 Supplement
+        - PIQRAY/Alpelisib
+        - Vitamin K
+        - Quadriplex
+        - Colchicine
+        - Thyro-Tabs
+        - Alprazolam/Xanax
+        - Spironolactone
+        - Vetstarch
+        - Enoxaparin
+        - Diclofenac/Voltaren
+        - Routin
+        - Doxepin
+        - Erythromycin
+        - Keterolac
+        - Tromethamine
+        - Cyclosporine/Atopica
+        - Pantoea agglomerans
+        - Oxybutynin
+        - Amikacin
+        - Levemir
+        - Apocaps
+        - Life Gold
+        - Red Clover Blossoms powder
+        - Modified Citrus Pectin
+        - Epinephrine
+        - Vitamin C
+        - Azathioprine
+        - RBC Transfusion
+        - Bactrim/sulfamethoxazole & trimethoprim
+        - Pet ReLeaf
+        - NeoPolyBac Ophthalmic
+        - Antibiotics (unspecified)
+        - Azithromycin
+        - Alendronate
+        - Cafazolin
+        - Diltiazem
+        - Mexiletine
+        - Pure IP6
+        - VetInsulin
+        - Herbal Supplements
+        - San Qi Formula
+        - Amnivast
+        - Crananidin
+        - Movoflex
+        - Lidocaine
+        - Tamsulosin/Flowmax
+        - Bedinvetmab/Librela
+        - Calcium Carbonate/Tums
+        - Dermatrophin
+        - Temozolomide
+        - Midazolam
+        - Anipryl
+        - Theophylline
+        - Sodium Bicarbonate
+        - RenaKare
+        - Hydroxazine
+        - Zincard/Dexrazoxane
+        - Animax
+        - Pro-Pectalin
+        - Ellevet CHews
+        - Cordyceps
+        - Benadryl
+        - Albon
+        - Robenacoxib (Onsior)
+        - Lysine
+        - Myos muscle building supplement
+        - Iron injections
+        - Xyzal (L-Cefirizine)
+        - Clavacillin
+        - Loperamide
+        - Theracurmin
+        - Quercetin Phytosome
+        - Anti-Neoplasia
+        - Fiber Supplement
+        - Zinc Supplement
+- surgery
+    - Whether surgical resection of the tumor was performed.
+    - For example:
+        <surgery>
+            <reasoning>[REASONING]</reasoning>
+            <resection>Yes</resection>
+        </surgery>
+- surgery_outcome
+    - The outcome of the surgery.
+    - For example:
+        <surgery_outcome>
+            <reasoning>[REASONING]</reasoning>
+            <outcome>Complete Resection</outcome>
+        </surgery_outcome>
+    - The following are the possible outcomes of the surgery:
+        - Completely Excised
+        - Incompletely Excised
+        - Unknown
+- metastasis_at_time_of_diagnosis
+    - Whether the cancer has spread to other parts of the body.
+    - For example:
+        <metastasis_at_time_of_diagnosis>
+            <metastasis>Yes</metastasis>
+        </metastasis_at_time_of_diagnosis>
+    - The following are the possible outcomes of the surgery:
+        - Yes
+        - No
+        - Unknown
+
+- compounding_pharmacy
+    - If a compounding pharmacy is listed, extract the name of the pharmacy. Do not include "fidocure" as a pharmacy name.
+    - For example:
+        <compounding_pharmacy>
+            <reasoning>[REASONING]</reasoning>
+            <pharmacy>CVS Pharmacy</pharmacy>
+        </compounding_pharmacy>
+
+- adverse_effects
+    - Any adverse effects the patient has experienced from the medications. For each adverse effect, extract the following fields:
+        - The name of the medication
+        - The dosage of the medication
+        - The date the adverse effect started
+        - A description of the adverse effect
+    - For example:
+        <adverse_effects>
+            <reasoning>[REASONING]</reasoning>
+            <medication>Doxorubicin/Adriamycin</medication>
+            <dosage>20 mg/kg</dosage>
+            <date>2021-01-01</date>
+            <description>Nausea</description>
+        </adverse_effects>
+
+- date_of_death
+    - The date of death of the patient, if it is known.
+    - For example:
+        <date_of_death>
+            <reasoning>[REASONING]</reasoning>
+            <date>2021-01-01</date>
+        </date_of_death>
+
+- weight
+    - The weight of the patient, if it is known. Convert all weights to kilograms.
+    - For example:
+        <weight>
+            <reasoning>[REASONING]</reasoning>
+            <weight>20 kg</weight>
+        </weight>
+"""

requirements.txt

@@ -0,0 +1,3 @@
+gradio
+openai
+pandas

run_extraction.py

@@ -0,0 +1,202 @@
+import glob
+import json
+import os
+import time
+import gradio as gr
+from openai import OpenAI
+
+import xml.etree.ElementTree as ET
+import re
+import pandas as pd
+import api_keys
+
+import note_extraction.hf_hosting.prompts as prompts
+
+client = OpenAI(api_key=api_keys.OPENAI_API_KEY)
+
+model_name = "gpt-4o-2024-08-06"
+
+demo = client.beta.assistants.create(
+    name="Information Extractor",
+    instructions="Extract information from this note.",
+    model=model_name,
+    tools=[{"type": "file_search"}],
+)
+
+def parse_xml_response(xml_string: str) -> pd.DataFrame:
+    """
+    Parse the XML response from the model and extract all fields into a dictionary,
+    then convert it to a pandas DataFrame with a nested index.
+    """
+    # Extract only the XML content between the first and last tags
+    xml_content = re.search(r'<.*?>.*</.*?>', xml_string, re.DOTALL)
+    if xml_content:
+        xml_string = xml_content.group(0)
+    else:
+        print("No valid XML content found.")
+        return pd.DataFrame()
+
+    try:
+        root = ET.fromstring(xml_string)
+    except ET.ParseError as e:
+        print(f"Error parsing XML: {e}")
+        return pd.DataFrame()
+
+    result = {}
+
+    for element in root:
+        tag = element.tag
+        if tag in ['patient_name', 'date_of_birth', 'sex', 'weight', 'date_of_death']:
+            result[tag] = {
+                'reasoning': element.find('reasoning').text.strip() if element.find('reasoning') is not None else None,
+                **{child.tag: child.text.strip() if child.text else None 
+                   for child in element if child.tag != 'reasoning'}
+            }
+        elif tag in ['traditional_chemo', 'other_cancer_treatments', 'other_conmeds']:
+            if tag not in result:
+                result[tag] = []
+            reasoning = element.find('reasoning')
+            for item in element:
+                if item.tag in ['drug', 'treatment', 'medication']:
+                    date_element = element.find('date')
+                    result[tag].append({
+                        'reasoning': reasoning.text.strip() if reasoning is not None else None,
+                        'name': item.text.strip() if item.text else None,
+                        'date': date_element.text.strip() if date_element is not None and date_element.text else None
+                    })
+        elif tag in ['surgery', 'surgery_outcome', 'metastasis_at_time_of_diagnosis']:
+            result[tag] = {
+                'reasoning': element.find('reasoning').text.strip() if element.find('reasoning') is not None else None,
+                **{child.tag: child.text.strip() if child.text else None 
+                   for child in element if child.tag != 'reasoning'}
+            }
+        elif tag == 'compounding_pharmacy':
+            result[tag] = {
+                'reasoning': element.find('reasoning').text.strip() if element.find('reasoning') is not None else None,
+                'pharmacy': element.find('pharmacy').text.strip() if element.find('pharmacy') is not None else None
+            }
+        elif tag == 'adverse_effects':
+            if tag not in result:
+                result[tag] = []
+            effect = {
+                'reasoning': element.find('reasoning').text.strip() if element.find('reasoning') is not None else None
+            }
+            for child in element:
+                if child.tag != 'reasoning':
+                    effect[child.tag] = child.text.strip() if child.text else None
+            if effect:
+                result[tag].append(effect)
+
+    # Convert to nested DataFrame
+    df_data = {}
+    for key, value in result.items():
+        if isinstance(value, dict):
+            for sub_key, sub_value in value.items():
+                df_data[(key, '1', sub_key)] = [sub_value]
+        elif isinstance(value, list):
+            for i, item in enumerate(value):
+                for sub_key, sub_value in item.items():
+                    df_data[(key, f"{i+1}", sub_key)] = [sub_value]
+        else:
+            df_data[(key, '1', '')] = [value]
+
+    # Create multi-index DataFrame
+    df = pd.DataFrame(df_data)
+    df.columns = pd.MultiIndex.from_tuples(df.columns)
+    
+    return df
+
+def get_response(prompt, file_id, assistant_id):
+    thread = client.beta.threads.create(
+        messages=[
+            {
+                "role": "user",
+                "content": prompts.info_prompt,
+                "attachments": [
+                    {"file_id": file_id, "tools": [{"type": "file_search"}]}
+                ],
+            }
+        ]
+    )
+    run = client.beta.threads.runs.create_and_poll(
+        thread_id=thread.id, assistant_id=assistant_id
+    )
+    messages = list(
+        client.beta.threads.messages.list(thread_id=thread.id, run_id=run.id)
+    )
+    assert len(messages) == 1
+    message_content = messages[0].content[0].text
+    annotations = message_content.annotations
+    for index, annotation in enumerate(annotations):
+        message_content.value = message_content.value.replace(annotation.text, f"")
+    return message_content.value
+
+def process(file_content):
+    if not os.path.exists("cache"):
+        os.makedirs("cache")
+    file_name = f"cache/{time.time()}.pdf"
+    with open(file_name, "wb") as f:
+        f.write(file_content)
+
+    message_file = client.files.create(file=open(file_name, "rb"), purpose="assistants")
+
+    response = get_response(prompts.info_prompt, message_file.id, demo.id)
+    df = parse_xml_response(response)
+    
+    if df.empty:
+        return "<p>No valid information could be extracted from the provided file.</p>"
+
+    # Transpose the DataFrame
+    df_transposed = df.T.reset_index()
+    df_transposed.columns = ['Category', 'Index', 'Field', 'Value']
+    df_transposed = df_transposed.sort_values(['Category', 'Index', 'Field'])
+
+    # Convert to HTML with some basic styling
+    html = df_transposed.to_html(index=False, classes='table table-striped table-bordered', escape=False)
+    
+    # Add some custom CSS for better readability
+    html = f"""
+    <style>
+    .table {{
+        width: 100%;
+        max-width: 100%;
+        margin-bottom: 1rem;
+        background-color: transparent;
+    }}
+    .table td, .table th {{
+        padding: .75rem;
+        vertical-align: top;
+        border-top: 1px solid #dee2e6;
+    }}
+    .table thead th {{
+        vertical-align: bottom;
+        border-bottom: 2px solid #dee2e6;
+    }}
+    .table tbody + tbody {{
+        border-top: 2px solid #dee2e6;
+    }}
+    .table-striped tbody tr:nth-of-type(odd) {{
+        background-color: rgba(0,0,0,.05);
+    }}
+    </style>
+    {html}
+    """
+    
+    return html
+
+def gradio_interface():
+    upload_component = gr.File(label="Upload PDF", type="binary")
+    output_component = gr.HTML(label="Extracted Information")
+
+    demo = gr.Interface(
+        fn=process,
+        inputs=upload_component,
+        outputs=output_component,
+        title="Clinical Note Information Extractor",
+        description="This tool extracts key information from clinical notes in PDF format.",
+    )
+    demo.queue()
+    demo.launch()
+
+if __name__ == "__main__":
+    gradio_interface()