As an intermediate phenotype for multiple cardiovascular diseases, left ventricular hypertrophy (LVH) benefits from early diagnosis, which allows for timely intervention to prevent worsening of the condition, mitigate severe complications like heart failure and arrhythmias, and consequently improve patient outcomes. Preliminary advances have been made using deep learning for the early diagnosis and identification of etiology in LVH. This paper reviews the pathophysiology, causes, and diagnostic standards for LVH, discusses the strengths and weaknesses of applying deep learning to diagnostic tools such as echocardiography, cardiac magnetic resonance imaging, and electrocardiogram, examines its use in prognostic evaluation, and concludes by summarizing current achievements and suggesting future research avenues.
Valvular heart disease (VHD) ranks as the third most prevalent cardiovascular disease, following coronary artery disease and hypertension. Severe cases can lead to ventricular hypertrophy or heart failure, highlighting the critical importance of early detection. In recent years, the application of deep learning techniques in the auxiliary diagnosis of VHD has made significant advancements, greatly improving detection accuracy. This review begins by introducing the etiology, pathological mechanisms, and impact of common valvular heart diseases. It then explores the advantages and limitations of using electrocardiographic signals, phonocardiographic signals, and multimodal data in VHD detection. A comparison is made between traditional risk prediction methods and large language models (LLMs) for predicting cardiovascular disease risk, emphasizing the potential of LLMs in risk prediction. Lastly, the current challenges faced by deep learning in this field are discussed, and future research directions are proposed.