Anticancer
Anticancer is an AI-powered dashboard for cancer treatment prediction, analyzing patient data to estimate cure rates and treatment duration.
Smarter Connections, Better Care!
Summary
Anticancer is an AI-powered dashboard that predicts cancer patients’ cure rates and treatment duration using big data. This project won the Grand Prize at the ‘AI & Data Mining for Cancer Big Data Competition’ hosted by the National Cancer Center of Korea.
As the team leader of a three-member team, I was responsible for data analysis, AI model development, and UI/UX dashboard design. To improve accuracy, we focused on handling missing values and enhancing UX visualization, ensuring usability for both medical professionals and patients.
My Role
All-Rounder
- Dashboard Design
- Data Analysis (SQL)
- AI development (python)
Team
3 members
- 3 developers of the ICT Convergence Department
Timeline
6-month Project
- Jun 30, 2021 → Dec 1, 2021
Tools
Figma
- Jira
- Notion
- VSCode
- Google Colab
Bridging the Gap: Improving Patient Understanding in Cancer Treatment
Effective doctor–patient communication is crucial. As shared decision-making becomes the norm, patients demand clearer, more accessible information.
1. When Words Create Barriers
"Many patients fail to accurately understand realistic treatment outcomes. This can lead to risky decisions, highlighting the need for improved doctor–patient communication."
2. Patients Perceive Their Condition More Optimistically Than Reality
Overly optimistic
37%
Misunderstand the treatment purpose
31%
Overestimate their chances of recovery
58.6%
What do they need?
- Clearer explanations are necessary.
- Minimize misinterpretation by visualizing complex medical information intuitively.
- Simplify terminology and provide intuitive graphs and treatment insights to enhance communication.
AI in Healthcare: Unlocking Data Potential, Supporting Doctors
AI is not here to replace doctors, but to assist them by reducing workload, improving diagnostic accuracy, and enhancing patient data analysis.
AI Can Assist, But Not Replace Doctors
lack patient face time
overwhelmed by patient demands
cite staff shortages
What tasks can AI assist with?
Medical Data Is Stored, But Rarely Used
So… What's the Solution?
- ✔ AI should reduce administrative burdens and unlock underutilized medical data.
- ✔ AI must support doctors with data-driven insights, enhancing decision-making.
- ✔ Bridge the gap between stored medical data and clinical application for better outcomes.
Doctor's Workflow with Anticancer
Patient List
1. Accessing Patient Data & Navigation
Patient Overview Panel
2. Reviewing Patient Information
Personalized Treatment Solutions
3. Evaluating AI Treatment Predictions
Data-Driven Patient Analysis
4. Comparing Similar Cases
AI Assessment Notes
5. AI-Assisted Clinical Insights
Doctor’s Notes
6. Finalizing & Documenting Treatment
Scheduling & Communication
7. Scheduling & Communication
Every Patient Detail at a Glance
The primary dashboard gives doctors immediate access to patient lists, daily summaries, and upcoming appointments. Personalized widgets highlight priority visits, real-time patient metrics, and relevant medical updates so clinicians can start each day fully informed.
- • Patient list sorted by urgency with quick access to visit details.
- • Summary cards surface vital stats, recent observations, and prescriptions.
- • Upcoming appointment calendar synchronizes with the hospital schedule.
Personalized Treatment Planning for Every Patient
The patient dashboard surfaces individualized care plans, AI predictions, and case comparisons in one view. Clinicians can adjust treatments, review AI assessment notes, and document decisions without leaving the screen.
- • Treatment success forecasts and risk levels update in real time.
- • Similar patient analytics offer evidence to support clinical decisions.
- • AI assessment notes and doctor’s notes keep communication aligned.
Patient Overview Panel
✔ Basic patient details & medical background
Automatically summarizes age, weight, smoking history, menopause status, and more so clinicians can understand the patient at a glance.
✔ Cancer classification & genetic markers
Displays tumor type (Ductal Carcinoma), BRCA gene mutations, and staging (T, N, M) in a structured, editable format.
💡 Why it matters?
Doctors get a quick snapshot of the patient's condition before making treatment decisions, ensuring nothing critical is overlooked.
Personalized Treatment Solutions
✔ Predicted treatment duration
AI estimates the expected treatment timeline based on patient history, helping doctors plan realistic care paths.
✔ Predicted cure rate
Displays cure probability so clinicians can compare therapeutic options with transparent risk profiles.
✔ AI-powered treatment recommendations
- Suggests the best treatment combinations using historical outcome data.
- Highlights potential side effects and recovery times for informed decision-making.
💡 Why it matters?
Clinicians can quickly evaluate AI predictions, compare options, and tailor treatment plans without switching views.
Similar Patient Cases
✔ Filterable comparisons
Toggle tumor staging, genetic markers, and treatment duration to see how comparable patients responded to care.
✔ Similarity scoring
Provides a clear similarity percentage to build confidence in selecting AI-suggested plans.
✔ Outcome insights
Review surgeries, treatments, and recovery outcomes from matching cases to guide final decisions.
💡 Why it matters?
Doctors can confirm treatment plans by studying proven outcomes from highly similar patients.
AI Assessment Notes
✔ Treatment duration & cure predictions
AI summarizes expected recovery timelines and cure probabilities, giving clinicians confidence in proposed plans.
✔ Alternative recommendations
When conditions worsen, the panel suggests escalation paths such as mastectomy or additional chemotherapy.
✔ Patient similarity analysis
Surfaces similar past cases with their outcomes to inform nuanced treatment adjustments.
💡 Why it matters?
AI notes reduce manual paperwork and keep every recommendation transparent, so doctors can justify treatment changes instantly.
Doctor's Notes
✔ Plan documentation
Clinicians record the agreed treatment plan, including hormone therapy and radiation regimens.
✔ Risk and monitoring guidance
Notes capture risk factors, follow-up schedules, and alternative options should the tumor progress.
✔ Patient lifestyle coaching
Final advice encourages healthy habits and proactive symptom reporting between visits.
💡 Why it matters?
Keeping physician notes inside Anticancer ensures every treatment decision, risk note, and patient instruction stays synchronized with the dashboard.
Big Data Processing & AI Implementation
Data Preprocessing
01
Outlier & missing-value handling
- ‘99: Not Applicable’ values replaced with NULL or mean imputation.
- Features like marriage count, age at first birth, and number of births standardized.
- Advanced imputation methods (KNN, MICE) prepared for robust handling.
Normalization
02
Scaling for model performance
- MinMaxScaler applies consistent scaling across features.
- Preprocessing removes outliers prior to scaling to preserve distributions.
- Ensures AI models converge faster with more stable gradients.
Feature Selection & Baseline Modeling
Feature Importance Analysis
03Gradient boosted trees surfaced the top 30 features influencing treatment outcomes, including tumor staging, hormone therapy, and BRCA markers.
Model Selection Workflow
04Jupyter notebooks iterate through baseline logistic regression, train/validation splits, and metric tracking to compare model performance.
- Train/test split (15%) with additional validation set for hyperparameter tuning.
- Baseline accuracy benchmarked against majority class before advanced models.
- Scikit-learn pipeline standardizes feature engineering for reproducibility.
Color System
Primary Color · 01
Dark Blue
HEX · #4F5C88
RGB · 79 | 92 | 136
Primary Color · 02
Blue
HEX · #0090FF
RGB · 0 | 144 | 255
Primary Color · 03
Light Blue
HEX · #C3DEF4
RGB · 195 | 222 | 244
Primary Color · 04
Light Pink
HEX · #F5D5E4
RGB · 245 | 213 | 228
