Dr. Imran Haque Has Some Thoughts on Deep-Learning Algorithms for Disease Diagnosis



Medical advances have grown alongside technological innovations, and researchers have leveraged deep learning and machine learning to create algorithms that can diagnose diseases and illnesses. A Stanford University researcher, Sebastian Thrun and his team created an algorithm that can identify whether any skin damage or blemishes are potentially cancerous, and the results match are as accurate as a board-certified physician. The AI introduced by Thrun is one of many algorithms that looks to help individuals diagnose illnesses through picture, x-rays, or facial recognition.

Dr. Imran Haque, an internist and general practitioner based out of North Carolina, is particularly excited about the benefits that these AI will have in the future. Specifically, he looks forward to a future where patients can leverage these computerized diagnostic processes so that they have targeted questions when meeting with a physician. Dr. Imran Haque will be using his extensive experience of over 15 years to help explain why a better-informed patient means for more precise and effective medicine.

The Black Box of Diagnostic Algorithms

When it comes to diagnostic algorithms, they learn as they are fed an incredible amount of information through various mediums – such as x-rays, photos of symptoms, or even facial features. By crunching the information, the algorithms are able to see correlations and linkages between certain features and illnesses to make diagnosis that human doctors cannot hope to achieve. As more data points are fed into the algorithm, the more accurate it will become over time. However, one of the important points is that it is sometimes unknown how and why the algorithm jumps to a specific conclusion. Dr. Imran Haque says that this is not something to worry about, as there is existing treatment that “works” even though the root cause is still unknown. For example: Lithium is approved to treating bi-polar disorder, but people still don’t know why it works.

A companion tools for doctors

An ultra-accurate computer does not mean that the job of doctors is replaced. Instead, Dr. Imran Haque sees diagnostic algorithms as a companion tool for doctors, much like the stethoscope. The tool will be able to provide an initial diagnosis While FDA approved algorithms need to be proven to be as good or better than the comparable physician, it’s still important to realize that mistakes could be a possibility. This is why the diagnostic algorithm will not be a replacement for doctors in the near future, but will instead be used as a companion tool by doctors.  The tool will make assumptions that are often accurate, but the doctor still needs to be available in order to verify the findings and administer the most effective treatment. In the end, the goal is for the AI tool to bolster the abilities of physicians, making for more precise diagnostics and treatments.

At-Home Diagnosis for Patients

Another benefit of the diagnostic algorithm is that some variations rely on information that is easy to provide, such as a photo of a symptom or a person’s face – which can make quick diagnosis easy and accessible for most consumers. This could lead to a future where consumers have access to some high-level diagnostic algorithms for home use as well. By providing the appropriate information for consumers, this will create a future where certain diseases can be identified much earlier and treated much quicker. Instead of waiting until a doctor’s appointment to get an opinion on something that may be cancerous, patients can have access to a regular scan – which can inform them if something looks amiss much earlier on in the lifecycle of potential cancer. This will also allow for patients to have targeted questions when they have appointments with physicians. Dr. Imran Haque points out that catching an illness or disease earlier on in the lifecycle means treatment will be both cheaper and easier to administer – as side effects are less severe and many more treatment options are available.

Sebastian Thrun’s Algorithm – The Future of Dermatology

When looking towards the future with Mr. Thrun’s potential skin cancer algorithm, it becomes clear that introducing the algorithm to consumers could create more demand for dermatologists in the near future. By making available an easy-to-use tool that provides accurate diagnosis to consumers, dermatologist will be called on more to verify the algorithms findings and potentially catch cancer earlier on in the lifecycle. Dr. Imran Haque firmly believes that medicine will improve if patients are more educated about their body and have a more targeted approach to appointments. Framing appointments around a precise question and hypothesis means that doctors can make more effective and efficient diagnosis as well, as time will not be wasted on so many exploratory questions.

At-Home Diagnosis for Consumers

Dr. Imran Haque knows that a future with better-informed consumers is a step forward in more efficient and effective medicine. The ability for algorithms to learn and give diagnosis based on basic information means that it will be possible for consumers to get very accurate diagnosis from their home  instead of relying on second-hand or third-hand information that might not be correct. In addition, the deep learning algorithms will only become more accurate as their sample size becomes larger. While there is concern that the “black-box” nature of these algorithms may make them unreliable, it’s important to realize that similar black-box treatments already exist. The diagnostic algorithms present a future where at-home diagnosis is more accurate and patients are better informed, thus making it easier for doctors to provide effective and efficient treatment.


Please enter your comment!
Please enter your name here