PD Insight Studio

A desktop application for Parkinson's disease real-world data analysis.

Built for dissertation-grade research — accepts any dataset, produces meaningful insights, comparisons, and optional predictions — all inside the app. No SPSS required.

---

What It Does

Tab	Purpose
① Load Data	Load one or more CSVs, preview, auto-normalise column names
② Configure	Map columns to roles (patient ID, time, group, outcome), build derived-variable rules, set QC thresholds
③ QC & Clean	Transparent quality control — every flagged row shows why it was flagged
④ Insights	Descriptive stats, group comparisons (effect sizes), time trends, progression analysis, narrative report
⑤ Predict	Exploratory classification or regression with feature importance
⑥ Export	Save cleaned CSV, markdown report, plots, and model results

---

How to Run in VS Code on Windows

1. Prerequisites

Make sure you have Python 3.11 installed:

python --version

Should print Python 3.11.x. If not, download from https://python.org.

---

2. Clone / Extract the Project

Place the pd_insight_studio folder somewhere on your machine, e.g.:

C:\Users\YourName\Documents\pd_insight_studio

---

3. Open in VS Code

File > Open Folder > select pd_insight_studio

---

4. Create a Virtual Environment

Open the VS Code Terminal (Ctrl+`):

python -m venv .venv
.venv\Scripts\activate

You should see (.venv) in the terminal prompt.

---

5. Install Dependencies

pip install -r requirements.txt

This installs pandas, matplotlib, seaborn, scikit-learn, scipy, and other required packages.

Note: tkinter is included with standard Python on Windows — no separate install needed.

---

6. Run the App

python main.py

The PD Insight Studio window will open.

---

Typical Workflow with mPower Data

1. Load Data → Add your mPower tapping CSV(s). The app auto-normalises column names.

2. Configure →

   • Set Patient/Subject ID → e.g. healthcode or recordid
   • Set Time/Visit column → e.g. createdon or timestamp
   • Set Primary group variable → e.g. medtimingstring (medication timing)
   • Set Outcome → e.g. tap_count or derived tap rate

3. Rule Builder →

   • Add a Keyword Map Rule: Source = medtimingstring, map keywords:

     before,immediately before → BEFORE
     after,just after,a while after → AFTER
     none,na,no → NO_MEDS
     

     Output column: med_state
   • Add a Formula Rule: tap_count / duration_seconds → tap_rate

4. QC → Run QC to see how many rows pass/fail and why.

5. Insights → Click "Generate All Insights" to see:

   • Distribution of tap rate per med group
   • Group comparison: BEFORE vs AFTER vs NO_MEDS effect sizes
   • Progression over time if visit timestamps are mapped

6. Predict → Optionally run a classification (PD vs Control proxy) or regression (tap rate).

7. Export → Save cleaned data, plots, and reports to your chosen folder.

---

Using with PPMI Data

• entity_id → patno
• time_col → infodt or visdate
• group_col → cohort or sex
• outcome_col → updrs_totscore or similar

Add a Months Since Baseline rule to convert dates to numeric visit month.

---

Project Structure

pd_insight_studio/
├── main.py          # GUI entrypoint
├── standardise.py   # Column normalisation, type inference, time parsing
├── rules.py         # Rule builder — keyword map, formula, months-since-baseline
├── qc.py            # Quality control — transparent flagging
├── insights.py      # Descriptive stats, group comparisons, progression, narrative
├── plotting.py      # All matplotlib/seaborn figures
├── modeling.py      # Logistic/Ridge regression + Random Forest with metrics
├── requirements.txt
└── README.md

---

Requirements

pandas>=2.0.0
numpy>=1.24.0
matplotlib>=3.7.0
seaborn>=0.12.0
scipy>=1.10.0
scikit-learn>=1.3.0
pillow>=10.0.0
openpyxl>=3.1.0

---

Important Notes

• Not hard-coded to mPower: any CSV dataset works — you configure the column roles yourself.
• No silent deletions: QC flags rows with reasons; you decide what to keep.
• Predictions are exploratory: All modelling output is clearly labelled as having no clinical validity.
• Large files: The app uses threading for QC and modelling to avoid freezing the UI.

---

Troubleshooting

Problem	Fix
ModuleNotFoundError	Make sure .venv is activated and you ran pip install -r requirements.txt
Window doesn't open	Check Python version is 3.11; try python -m tkinter to verify Tkinter works
Plots don't show	Ensure matplotlib is installed; restart app
Large file is slow to load	This is normal for >100k rows. Wait for progress or split the file