Artificial intelligence is becoming a larger part of conversations about diagnosis, medical imaging, and patient care. Andrew Ting brings clinical experience to the discussion about how these systems should be developed and evaluated for real healthcare settings. That perspective matters because a tool that performs well during development still has to make sense when doctors use it with actual patients.
Start With Problems Clinicians Actually Face
Medical AI should make a doctor’s job easier or help solve a problem that affects patient care. Clinicians see firsthand where patients wait too long, where more information would help, and which routine tasks take up valuable time. Bringing that experience into development can keep teams focused on tools that doctors and patients may actually need.
Even a useful idea can cause problems if it does not fit naturally into a doctor’s day. Healthcare workers already move between patient records, test results, medical images, and other systems while caring for patients. If a new AI tool adds several more steps to that routine, doctors may spend more time managing the technology than benefiting from it.
Teach AI With Relevant Medical Data
An AI system learns from the information used during its development. In medicine, that information may include images, clinical measurements, laboratory findings, or other health data. The quality and relevance of those examples can strongly influence how useful the resulting system becomes.
Clinical knowledge helps teams recognize whether the data represent the patients who may eventually encounter the technology. A dataset that leaves out important patient groups or common variations may produce results that do not translate well into broader care. Developers therefore need to understand both the quantity of their data and what those records actually represent.
Give Medical Context to the Numbers
Strong test results do not always mean an AI system will be useful in a real clinic. Developers also need to think about the disease being evaluated and what could happen when the system gets something wrong. Missing signs of a serious condition, for example, can have very different consequences from sending a healthy patient for an extra examination.
This is where Andrew Ting MD experience can add a clinical point of view to the development process. Doctors understand which errors could put a patient at greater risk and which results should be checked more closely before anyone acts on them. Their input helps developers look beyond a promising accuracy score and pay attention to what those numbers could mean for an actual patient.
Look Closely at False Results
False positives and false negatives deserve careful attention when evaluating medical AI. A false positive may lead to more testing, worry, or unnecessary appointments, while a false negative could delay appropriate care. Neither result can be understood properly by looking at overall accuracy alone.
Clinical experience can help a development team decide which errors deserve the most attention for a particular application. The acceptable balance may differ between a screening tool and a system intended to support a more specific diagnosis. Understanding the intended use should therefore come before deciding what performance level is acceptable.
Build Tools Doctors Can Actually Use
An AI tool may work perfectly in testing and still be a headache for doctors to use. During a busy appointment, doctors don’t have time to type the same information twice, switch between several screens, or figure out what a complicated result means. Developers need to understand how the tool fits into the doctor’s normal routine before adding more steps to an already busy day.
The information should also be easy to understand when a doctor needs it. A clinician should be able to see what the system found and decide whether the result deserves a closer look. If the technology is confusing or slows everyone down, impressive performance numbers may not matter much once it reaches a real clinic.
Keep the Clinician in the Decision
AI can identify patterns quickly, but medical decisions involve information that may extend beyond what a model receives. Symptoms, medical history, medications, family history, physical findings, and conversations with the patient can all influence a doctor’s assessment. A useful system should support that broader evaluation rather than pretend to replace it.
Training developers to understand this distinction can prevent unrealistic expectations about automation. A prediction or alert may provide another piece of information for the clinician to consider. The final decision still needs appropriate medical context, particularly when the consequences for the patient are significant.
Protect Patient Information During Development
Building medical AI often means working with health records and other information patients expect to remain private. Before using that data, teams should know exactly what they need, who can see it, and how they will keep it secure. Teams should make those decisions from the beginning instead of waiting until the system is almost finished.
Doctors also understand why patients may feel uneasy about their medical information being used to develop new technology. A health record can reveal personal details that someone may not want shared beyond the people involved in their care. Developers need strong security measures and clear rules for how information is accessed, stored, and used.
Test Beyond the Development Environment
A system that performs well with development data still needs careful evaluation before its results can be trusted elsewhere. Different hospitals and clinics may serve different populations, use different equipment, or collect information differently. These changes can affect how a model performs after leaving the setting where it was created.
Testing across realistic conditions can reveal weaknesses that were not obvious earlier. Clinicians can help identify unusual cases, workflow problems, or patient differences that deserve closer attention. Their involvement gives developers a clearer view of how the technology behaves outside a controlled development environment.
Final Thoughts
Building better medical AI takes more than skilled developers and plenty of medical data. Andrew Ting brings a doctor’s perspective to questions about patient care, privacy, clinical decisions, and how new technology will actually fit into a busy healthcare setting. Keeping clinicians involved can help ensure these tools solve real problems instead of creating new ones for doctors and patients.
