ABHINAYA K(STM21CS005)
ANAMIKA K(STM21CS017)
GOPIKA PRADEEP(STM21CS033)
SNEHA AJITH(STM21CS060)

Cardiac diseases encompass a range of conditions affecting the heart’s structure and function, leading to significant morbidity and mortality worldwide. Early detection and accurate diagnosis of these conditions, including Severe Left Ventricular Hypertrophy (SLVH), Dilated Left Ventricle (DLV) and Ejection Fraction (EF) abnormalities, are crucial for effective management and improved patient outcomes. Traditional diagnostic methods often rely on single-modality data, which may overlook the multifaceted nature of cardiac health.
This paper proposes an advanced multimodal deep learning framework that integrates echocardiographic images with chest X-ray (CXR) structured data and CXR imagery to enhance the prediction accuracy of cardiac diseases. By leveraging the complementary strengths of these diagnostic modalities, the proposed system aims to provide a more holistic understanding of cardiac health, facilitating the early detection of SLVH, DLV, and EF assessment.
The innovative approach employs techniques such as Variational Autoencoders (VAEs) for data fusion, EfficientNetB3 for feature extraction, and attention mechanisms to focus on clinically relevant features. Additionally, a Decision Tree classifier is utilized for the final classification, ensuring an interpretable and effective diagnostic model. Furthermore, the system incorporates a treatment recommendation module that suggests appropriate therapeutic interventions based on the detected condition, aiding clinicians in
optimizing patient care.