Clinical Scorecard: Deep Learning-Based Integration of EHR and ECG for Anticipating New-Onset Coronary Heart Disease in Patients with Cancer

At a Glance

Key Highlights

• 722 out of 1262 patients (57.2%) developed new-onset CHD.
• CNN-LSTM model achieved an AUC of 0.975, outperforming seven conventional machine learning models.
• Subgroup analysis showed AUC of 0.924 in early-stage and 0.888 in late-stage patients.
• Strong predictive features included PAB, CRP, age, cTnI, PT, and ECG markers.

Guideline-Based Recommendations

Diagnosis

• New-onset CHD diagnosed based on current clinical guidelines.

Management

• Early identification of cancer-associated CHD is critical for improving prognosis.

Monitoring & Follow-up

• Continuous monitoring of biomarkers such as cTn and CRP.

Risks

• Cancer patients have a significantly higher risk of developing CHD due to treatments and underlying factors.

Patient & Prescribing Data

Cancer patients receiving chemotherapy, targeted therapy, immunotherapy, or radiotherapy.

Elevated cTn levels following treatment correlate with adverse outcomes.

Clinical Best Practices

• Utilize multimodal prediction approaches for cardiovascular risk assessment.
• Incorporate various biomarkers and clinical features in risk stratification.

Related Resources & Content

• Study on AI-driven models for cardiac risk stratification