As a veteran of multiple healthcare startups, it’s hard not to become immune to claims of technological developments which will “revolutionize patient care”.
To be fair, revolutionary changes do exist. Telehealth for example, enabled by the combination of AI and cloud technology, has revolutionized the access patients have to robust diagnostic services. Yet, telehealth with its data fragmentation and current clinical diagnostic limitations, still has serious challenges making remote diagnostic interventions a secondary modality for many. That, I believe, is about to change with the development of Multimodal AI.
Something I believe which qualifies as a true revolutionary development.
Why Telehealth As It Currently Exists, Isn’t Good Enough Anymore
Healthcare systems around the world are stressed. Driven largely by staff shortages and infrastructure limitations, access to diagnostic healthcare services has been especially challenging for those in rural and lower income communities and countries. It’s not uncommon for patients in these settings to wait weeks or even months to access the most basic diagnostic services.
Fortunately, telehealth, with its ability to connect remote patients with radiologists, has significantly improved access to diagnostic care for millions. Something we should be grateful for.
But let’s not kid ourselves. Telehealth as it currently exists, is far from perfect.Telehealth clinicians usually work without integrated lab results, imaging data, and longitudinal patient history, making accurate diagnoses difficult. This is true too for the traditional AI models supporting telehealth services, where only a single data type (e.g., images) is used to train the AI model. However, this is the biggest flaw that doctors and physicians have shown us: human diagnosis is a multimodal affair; it cannot be realized from just one source of data. A doctor doesn’t base their conclusions on a single datum, but rather combines imaging, history, physicals, and intuition to make a diagnosis.
We need to stop playing with fractured tools and move towards a holistic understanding of patients through systems.
That’s where multimodal AI can and should revolutionize the healthcare industry.
Multimodal AI Is a Clinical Force
Unlike traditional telehealth AI models, multimodel AI is a type of artificial intelligence (AI) modelling which integrates and interprets data from a variety of sources (e.g., text, images, audio and video). By leveraging data from medical imaging, electronic health records, wearable sensors, genomics and patient-reported symptoms, providers using multimodel AI technology are creating conditions clinically comparable to traditional care settings. An AI resource that is capable of digesting a varied array of data inputs and calculating for example, the likelihood of tumor progression based on the patient’s genetics, history and data on lifestyle. Armed with these insights, clinicians can more quickly and accurately triage patients.
A game changer for under-resourced areas with limited staff.
Powering the Virtual Hospital of the Future
More than this, multimodal AI brings us closer to realizing what was once just a pipe dream; the “virtual hospital”.
Imagine a cloud-based diagnostic platform that continuously receives patient inputs from a myriad of sources: blood pressure from a wearable, lab results from a local clinic, a chest scan uploaded from a remote hospital, and medication adherence via smart pill bottles, etc. With multimodal AI, a clinician thousands of miles away, is able to access all of this through a secure dashboard driven by AI that interprets it in real time, flags risk, and recommends a course of action. All within just a few minutes.
This idea is not a hypothetical. Select healthcare startups are moving in this direction. Companies such as Aidoc and PathAI for example, are already able to demonstrate the power of multimodal AI in niche diagnostic areas. This new wave of startups will step out of a single-use model and provide multimodal solutions with scalable cloud infrastructure, allowing healthcare systems and providers to offer superior care regardless of geography.
For Investors, the Opportunity Is Now
Investment-wise, the next several years will be a definitive time for healthcare AI. The AI in the medical imaging market alone is expected to hit $8.18 billion by 2030, with a CAGR of almost 35%, according to Grand View Research. Combine that with wearables, lab tech and EHR integrations, the potential for value creation has never been bigger.
However, we cannot treat AI startups equally. For multimodal AI to work, a company requires a very different set of resources:
- Access to large, high-quality, diverse datasets
- Partnerships with healthcare providers
- Cloud architecture that ensures security and scalability
- A clear roadmap for FDA approval or international regulatory clearance
Furthermore, start-ups need to focus on equity and inclusiveness in their design. Algorithmic bias in healthcare AI is not a theoretical anxiety but a real threat. A virtual hospital that does not take into consideration an array of factors such as gender, racial, and geographic diversity in its training data, will only exacerbate the disparities we are trying to solve.
Building a Future That’s Both Digital and Human
From my vantage point as a healthcare startup veteran, multimodal AI represents the most significant technological advancement digital health has experienced since the advent of the EHR. Nevertheless, we have to be clear about its role: AI is not here to replace clinicians. It’s a supportive tool to expand access, decrease their workload and share their knowledge with the masses.
Our work as investors, builders, and providers is to make these systems transparent, fair, and human. If we do succeed, we will not only improve health care, we will redefine what it means to have access to care.
About Thomas Kluz
Thomas Kluz is a distinguished venture capitalist with over a decade of experience. He’s the Managing Director of Niterra Ventures, where his investments focus on energy, mobility, and healthcare. With deep expertise in healthcare-focused venture capital, he has a proven track record of success with various organizations, such as Qualcomm Ventures and Providence Ventures.
