Abstract 4373117: Quantum Computing based Echocardiographic Diagnosis and Analysis in Congenital Heart Disease: Feasibility and Superiority to conventional Deep Learning Approaches

Background: Quantum computing (QC) has emerged as an innovative technology to enhance machine learning through quantum mechanical properties such as superposition, entanglement, and projection into high-dimensional complex Hilbert spaces. These properties can improve the representational capacity and generalization of deep learning models. We explore the feasibility and efficacy of integrating quantum technology into convolutional neural networks (CNN) for echocardiographic analysis of real world data covering the spectum of congenital heart disease (CHD). Methods: We developed a hybrid deep learning algorithm incorporating a quantum computing (QC) layer within a ResNet-50 convolutional neural network. The QC layer included 4 qubits, with angle embedding and strongly entangling layers for quantum processing. Two supervised classification tasks were evaluated: (1) diagnosis classification and (2) echocardiographic view classification using a challenging dataset of echocardiographic images from patients with congenital and structural heart disease. The model was benchmarked against a conventional state-of-the art CNN model. Training/inference were conducted on high-performance classical GPUs and the QC layer were simulated and applied directly to an IBM 127-qubit Eagle r3 quantum processor (Sherbrooke, Canada). Results: Models were trained and tested on echocardiographic data derived from 262 patients and 62 controls. Diagnoses included tetralogy of Fallot (n=30), TGA (n=48), Ebstein anomaly (n=18), and other CHD/structural anomalies. A total of 9,793 loops including 284,250 frames were used for training and testing.The hybrid QC-model demonstrated superior performance in both tasks relative to the classical baseline, with improvements in accuracy, F1-score, precision, and recall. Per-class metrics showed enhanced differentiation in diagnostically challenging categories (Test accuracy 72.1% vs. 68.4% for diagnosis and 78.9 vs. 76.6% for view classification for QC vs. conventional CNN, respectively). Conclusion: This study is first to establish the applicability of quantum-computing deep learning in the field of congenital cardiology. The enhanced ability of quantum layers to map complex image data into higher-dimensional spaces offers promising advantages for AI applications, particularly in populations with high anatomic variability such as CHD.The findings pave the way for future quantum applications in precision medicine specifically benefiting CHD.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.