Alat AI Diagnos Kegagalan Jantung Lanjut dengan Ketepatan Tinggi

Merevolusikan Diagnosis Kegagalan Jantung Lanjut dengan AI

Kegagalan jantung lanjut, satu keadaan yang melemahkan yang menjejaskan ratusan ribu orang di seluruh dunia, telah lama menimbulkan cabaran diagnostik yang signifikan. Pesakit sering mengalami kelewatan diagnosis disebabkan oleh sifat kaedah penilaian semasa yang kompleks dan memerlukan sumber yang intensif. Walau bagaimanapun, satu kajian terobosan daripada pasukan kolaboratif di Weill Cornell Medicine, Cornell Tech, Cornell Ann S. Bowers College of Computing and Information Science, Columbia University Vagelos College of Physicians and Surgeons, dan NewYork-Presbyterian bersedia untuk mengubah landskap ini. Penyelidik telah berjaya membangunkan dan menguji alat yang dikuasakan kecerdasan buatan (AI) yang boleh mengenal pasti pesakit dengan kegagalan jantung lanjut dengan ketepatan tinggi menggunakan data ultrasound jantung rutin dan rekod kesihatan elektronik (EHR). Pendekatan inovatif ini menjanjikan untuk mendemokrasikan diagnosis dan meningkatkan penjagaan pesakit secara signifikan.

Kekangan Diagnostik: Mengapa AI Kritikal

Pada masa kini, diagnosis muktamad kegagalan jantung lanjut sangat bergantung pada ujian senaman kardiopulmonari (CPET). Walaupun berkesan, CPET adalah prosedur khusus yang memerlukan peralatan mahal dan kakitangan yang terlatih tinggi, menjadikannya hanya tersedia di pusat perubatan akademik yang besar. Ini mewujudkan kekangan diagnostik yang besar, menyebabkan anggaran 200,000 rakyat Amerika dengan kegagalan jantung lanjut kurang mendapat perkhidmatan atau tidak didiagnos setiap tahun. Kekurangan akses meluas kepada CPET bermakna ramai pesakit terlepas peluang untuk intervensi tepat pada masanya dan penjagaan khusus.

Kaedah baharu yang dikuasakan AI ini secara langsung menangani isu ini dengan menyediakan penyelesaian diagnostik yang lebih mudah diakses dan boleh diskala. "Ini membuka laluan yang menjanjikan untuk penilaian pesakit dengan kegagalan jantung lanjut yang lebih cekap menggunakan sumber data yang sudah tertanam dalam penjagaan rutin," jelas Dr. Fei Wang, associate dean untuk AI dan sains data dan Frances and John L. Loeb Professor of Medical Informatics di Weill Cornell Medicine, dan penulis kanan kajian. Dengan meramalkan penggunaan oksigen puncak (peak VO2) — ukuran CPET yang paling kritikal — daripada imej ultrasound dan data EHR yang mudah diperoleh, model AI mengelak kekangan tradisional, memastikan lebih ramai pesakit dapat dikenal pasti dan menerima penjagaan yang sesuai.

Keupayaan luar biasa alat AI ini berpunca daripada model pembelajaran mesin berbilang modal, berbilang contoh yang canggih. Dibangunkan oleh pasukan Dr. Wang, termasuk penulis utama Dr. Zhe Huang dan Dr. Weishen Pan, model ini boleh memproses beberapa jenis data yang berbeza secara serentak, menawarkan pandangan komprehensif tentang kesihatan jantung pesakit.

Jenis Data	Penerangan	Peranan dalam Model AI
Ultrasound Bergerak Biasa	Imej dinamik yang menunjukkan struktur dan fungsi jantung	Isyarat visual untuk kontraktiliti jantung, saiz bilik, dan pergerakan dinding
Imej Bentuk Gelombang	Perwakilan grafik dinamik injap jantung dan corak aliran darah	Pandangan tentang anomali aliran darah dan fungsi injap
Rekod Kesihatan Elektronik	Demografi pesakit, sejarah perubatan, keputusan makmal, ubat-ubatan, dsb.	Maklumat kontekstual untuk profil pesakit yang holistik

Keupayaan untuk menggabungkan dan mentafsir pelbagai aliran data ini membolehkan AI mempelajari corak kompleks yang menunjukkan kegagalan jantung lanjut yang mungkin terlepas melalui analisis data terpencil. Model ini dilatih dengan teliti menggunakan data yang tidak dikenal pasti daripada 1,000 pesakit kegagalan jantung di NewYork-Presbyterian/Columbia University Irving Medical Center. Selepas latihan, prestasinya disahkan pada kohort baharu seramai 127 pesakit kegagalan jantung dari tiga kampus NewYork-Presbyterian yang lain. Keputusan adalah meyakinkan, menunjukkan ketepatan keseluruhan kira-kira 85% dalam membezakan pesakit berisiko tinggi. Ketepatan tinggi ini menunjukkan potensi kegunaannya dalam tetapan klinikal dunia sebenar, menawarkan penanda aras baharu untuk menilai agen AI untuk pengeluaran dalam diagnostik perubatan.

Keputusan Menjanjikan dan Inovasi Kolaboratif

Kejayaan alat AI ini adalah bukti kepada kekuatan kolaborasi antara disiplin, satu ciri inisiatif AI Kardiovaskular, usaha yang lebih luas oleh Cornell, Columbia, dan NewYork-Presbyterian. Dr. Nir Uriel, pengarah kegagalan jantung lanjut dan transplantasi jantung di NewYork-Presbyterian, memainkan peranan penting dalam memulakan projek ini. "Pada awalnya kami mengumpulkan sekumpulan lebih daripada 40 pakar kegagalan jantung dan meminta mereka memberitahu kami di mana mereka fikir AI boleh diaplikasikan dengan terbaik," katanya. Pendekatan yang diterajui oleh pakar klinikal ini memastikan bahawa penyelesaian AI secara langsung menangani keperluan klinikal yang kritikal.

Dr. Deborah Estrin, associate dean untuk impak di Cornell Tech, menekankan hubungan simbiotik: "Interaksi rapat antara pakar klinikal dan penyelidik AI dalam projek ini akhirnya memacu pembangunan teknik AI baharu yang tidak akan diterokai sebaliknya. Jadi, ini adalah kes perubatan membentuk masa depan AI — bukan hanya AI membentuk masa depan perubatan." Semangat kolaboratif ini, yang menghubungkan kepakaran klinikal dengan penyelidikan AI canggih, adalah penting untuk membangunkan alat yang teguh dan relevan secara klinikal. Perkongsian sedemikian adalah penting untuk memajukan aplikasi AI dalam domain sensitif seperti penjagaan kesihatan, di mana privasi data dan pertimbangan etika adalah sangat penting. Usaha sekitar privasi perusahaan dalam mengendalikan data perubatan sentiasa berkembang.

Membuka Jalan untuk Integrasi Klinikal dan Impak Masa Depan

Keputusan yang menjanjikan daripada kajian ini menandakan langkah signifikan ke arah mengintegrasikan AI ke dalam penjagaan kardiovaskular rutin. Pasukan penyelidik sudah merancang kajian klinikal, satu fasa yang perlu untuk mendapatkan kelulusan Pentadbiran Makanan dan Dadah A.S. (FDA) dan penerimaan klinikal yang meluas selepas itu. Dr. Uriel menggariskan potensi transformatif: "Jika kita boleh menggunakan pendekatan ini untuk mengenal pasti ramai pesakit kegagalan jantung lanjut yang mungkin tidak dikenal pasti sebaliknya, maka ini akan mengubah amalan klinikal kita dan meningkatkan hasil pesakit serta kualiti hidup secara signifikan."

Alat AI ini mewakili lebih daripada sekadar kemajuan teknologi; ia adalah anjakan paradigma dalam cara kegagalan jantung lanjut boleh didiagnos, menjadikan perubatan ketepatan lebih mudah diakses. Dengan memanfaatkan infrastruktur sedia ada (mesin ultrasound) dan data yang tersedia secara meluas (EHR), model ini mengurangkan halangan kepada pengesanan awal, memastikan lebih ramai pesakit menerima rawatan tepat pada masanya yang menyelamatkan nyawa. Kejayaan inisiatif ini pasti akan menginspirasi penerokaan lanjut ke dalam peranan AI dalam pelbagai kepakaran perubatan, akhirnya meningkatkan ketepatan diagnostik dan penjagaan pesakit secara menyeluruh.

Sumber asal

https://news.weill.cornell.edu/news/2026/03/ai-tool-shows-promise-in-diagnosing-advanced-heart-failure

Soalan Lazim

What is advanced heart failure and why is its diagnosis challenging?

Advanced heart failure is a severe, chronic condition where the heart struggles to pump enough blood to meet the body's needs, significantly impacting quality of life and prognosis. Diagnosing this condition accurately is critically challenging due to its complex nature and the limitations of current standard diagnostic procedures. The gold standard, cardiopulmonary exercise testing (CPET), requires specialized equipment and highly trained personnel, making it accessible only in large, tertiary medical centers. This bottleneck means that a vast number of patients who could benefit from advanced therapies are often overlooked or diagnosed late, delaying crucial interventions and worsening outcomes. The difficulty in early and widespread detection underscores the urgent need for more accessible and efficient diagnostic methods, which this new AI tool aims to provide by simplifying the diagnostic pathway and democratizing access to timely identification of the condition.

How does the new AI tool specifically improve upon existing diagnostic methods like CPET?

The AI tool developed by Weill Cornell Medicine and its partners represents a significant leap forward by overcoming the inherent limitations of CPET. Unlike CPET, which demands specialized facilities and staff, the AI model utilizes readily available cardiac ultrasound images and electronic health records (EHRs)—data sources routinely collected in most clinical settings. By processing these common data types, the AI can predict peak oxygen consumption (peak VO2), the most crucial measure derived from CPET, with high accuracy. This dramatically reduces the need for expensive, time-consuming, and resource-intensive CPETs. The improvement lies in its scalability and accessibility; it transforms a complex diagnostic process into one that can be integrated into routine clinical care, potentially identifying tens of thousands more patients who would otherwise go undiagnosed due to geographical or resource constraints.

What types of data does the AI model leverage for its predictions?

The innovative AI model is a multi-modal, multi-instance machine learning system, designed to synthesize information from diverse clinical data sources for a comprehensive assessment. It specifically processes three distinct categories of data. Firstly, it analyzes ordinary moving ultrasound images of the heart, capturing critical visual information about cardiac structure and function. Secondly, it incorporates related waveform imagery, which displays intricate details of heart valve dynamics and blood flow patterns. Lastly, the model integrates various items found in the patient’s electronic health records (EHRs), including demographic information, medical history, lab results, and other clinical parameters. The ability to combine and interpret these disparate data types allows the AI to develop a holistic understanding of a patient's cardiac health, leading to more accurate predictions.

What was the accuracy of the AI model in predicting peak VO2, and what does this mean clinically?

The AI model achieved a remarkable overall accuracy of roughly 85% in predicting peak VO2, which is a significant indicator for distinguishing high-risk patients with advanced heart failure. This figure was measured using a metric that assesses the probability of a randomly chosen high-risk patient having a higher predicted risk than a randomly chosen lower-risk patient. Clinically, an 85% accuracy suggests that the tool is highly effective and reliable in identifying individuals who require advanced heart failure care. This level of precision means that the AI can act as a powerful screening or supplementary diagnostic tool, helping clinicians more confidently and quickly identify patients who would benefit most from further evaluation or specialized treatments. This promising result paves the way for potential FDA approval and widespread adoption in healthcare.

Which institutions and key individuals collaborated on the development of this AI tool?

This groundbreaking study was the result of a highly collaborative effort involving several leading institutions and prominent experts. Key collaborating entities included Weill Cornell Medicine, Cornell Tech, Cornell Ann S. Bowers College of Computing and Information Science, Columbia University Vagelos College of Physicians and Surgeons, and NewYork-Presbyterian. The study was senior-authored by Dr. Fei Wang, Associate Dean for AI and Data Science at Weill Cornell Medicine. Other pivotal contributors included Dr. Deborah Estrin, Associate Dean for Impact at Cornell Tech, and Dr. Nir Uriel, Director of Advanced Heart Failure and Cardiac Transplantation at NewYork-Presbyterian. The AI team, under Dr. Wang, also included lead authors Dr. Zhe Huang and Dr. Weishen Pan, along with students and faculty from Cornell Bowers, highlighting a robust interdisciplinary approach to medical innovation.

What are the next steps for bringing this AI diagnostic tool into routine clinical practice?

The research team is actively planning the crucial next steps required to transition this promising AI diagnostic tool from a research finding to routine clinical practice. The immediate focus will be on conducting extensive clinical studies. These studies are essential to further validate the model's performance in diverse patient populations and real-world clinical settings, gathering the robust evidence necessary for regulatory approvals. Achieving U.S. Food and Drug Administration (FDA) approval is a critical milestone for widespread adoption. Following approval, efforts will concentrate on integrating the AI tool seamlessly into existing healthcare workflows and electronic health record systems. The ultimate goal is to enable clinicians to easily leverage this technology, ensuring more patients with advanced heart failure are identified and receive appropriate care in a timely manner, transforming current diagnostic paradigms.

How does this research embody the intersection of medicine and AI innovation?

This research project serves as a prime example of how the synergy between medicine and AI innovation can drive transformative advancements in healthcare. It began with clinicians, specifically heart failure specialists, identifying a critical unmet need—the diagnostic bottleneck for advanced heart failure. This clinical challenge then inspired AI experts to develop novel machine learning techniques, demonstrating a unique 'medicine shaping AI' dynamic. The multi-modal AI model, capable of interpreting complex medical images and electronic health records, showcases AI's potential to extract subtle, actionable insights that might elude human analysis or standard tests. This interdisciplinary approach not only addresses a significant clinical problem but also pushes the boundaries of AI research, developing models specifically tailored for the intricacies of medical data and clinical decision-making. It highlights the power of collaborative innovation in solving real-world healthcare issues.

What are the broader implications of this AI tool for patient care and healthcare systems?

The broader implications of this AI diagnostic tool for patient care and healthcare systems are profound. Firstly, it promises to significantly improve patient outcomes and quality of life by enabling earlier and more accurate identification of advanced heart failure. This means patients can receive timely interventions, advanced therapies, or even transplantation referrals, preventing disease progression and reducing mortality. Secondly, it addresses health disparities by making advanced diagnostic capabilities accessible beyond specialized centers, potentially reaching underserved populations. For healthcare systems, the tool could lead to more efficient resource allocation, reducing the burden on CPET facilities and streamlining diagnostic pathways. It also sets a precedent for how AI can be integrated into routine care to augment clinical decision-making, offering a scalable solution to complex medical challenges and enhancing the overall precision and accessibility of cardiovascular medicine.

Kekal Dikemas Kini

Dapatkan berita AI terkini dalam peti masuk anda.

Kongsi