Bias in diagnosing spinal leaks, particularly in radiology, has impacted patient outcomes for a long time. Radiologists, despite their expertise, are still human, which means they can introduce bias when interpreting complex images like brain MRIs.
Such bias often leads to missed or dismissed symptoms, creating a significant financial and emotional burden on patients. For people with spinal leaks, failure to diagnose is costly to both the patients’ wallets and their quality of life. Some of these patients end up spending as much as $20,000 annually on conservative treatments like IV saline, which pose high infection risks, leading to the unfortunate deaths of multiple spinal leak patients from PICC line infections. Patients also spend extensively on unnecessary treatments in their desperation to have relief from their spinal leak–unnecessary treatments that sometimes result in further disability.
For instance, one pain management anesthesiologist was attempting to propose intrathecal prolotherapy (PRP) for headaches from a spinal leak when research has not covered the use of PRP. This approach does not follow the American Society of Anesthesiologists (ASA) guidelines and would have created yet another dural puncture and subsequent spinal leak, with research showing a rate of approximately 60% for spinal leak symptoms persisting past the 18-month point after a dural puncture.
The emergence of artificial intelligence (AI) offers a promising solution to improve the accuracy of cerebrospinal fluid (CSF) leak diagnoses while reducing diagnostic bias.
An overview of bias in healthcare diagnostics
Over 20 million people in the U.S. are estimated to have chronic spinal leaks due to failure to diagnose in the acute stages. This number is increasing rapidly each year and creating an immense disease burden on the American population, especially impacting women patients. It’s crucial to reduce diagnostic bias to ensure these patients receive the treatments they desperately need, like epidural blood patches.
Human error and bias in interpreting diagnostic imaging have been well documented, particularly regarding brain MRIs. Radiologists often work under immense pressure and sometimes overlook subtle indicators of spinal leaks.
For instance, tonsillar descent, a sign of low cerebrospinal fluid volume, is sometimes dismissed as unimportant when it’s less than 5 mm despite its undeniable connection to symptoms like headaches. It’s one of many signs that can point to intracranial hypovolemia secondary to a spinal leak.
However, many female patients have reported that male doctors told them that it is normal for women to have headaches or exaggerate their pain. This creates a problem where a radiologist might see the clinical indication of a headache and, thus, dismiss significant critical findings on a brain MRI due to downplaying women’s pain. Such biases highlight the limitations of relying on a single human assessor for diagnosing spinal leaks.
There are some concerns about artificial intelligence developing a bias from the data it accesses, but we would offer the idea that AI is easier to train for bias reduction than humans. A reduced-biased AI system offers a solution to counterbalance human subjectivity. Artificial intelligence in healthcare can help improve diagnostic accuracy in detecting spinal leaks.
The Bern Score and its role in diagnosing spinal leaks
The Bern Score is a quantitative, multi-feature method developed to analyze subtle shifts in brain structures visible on MRIs for signs of intracranial hypovolemia secondary to a spinal leak as a noninvasive assessment option. The Bern Score is an adaptable tool for diagnosing CSF hypovolemia, examining many different features that can shift rather than solely looking for one feature, as many radiologists do. This examination considers many different features and is more suitable for how patient biological diversity and spinal leak severity can cause drastically different symptoms.
Dr. Beck, one of the developers of the Bern Score, emphasizes the scoring system’s potential as a starting point and encourages future researchers to expand its criteria, using AI to not only administer the assessment but also to pick up on new features that prove to be reliable in showing hypovolemia that can be added to the scoring system.
Given that outdated methods like CT myelograms are invasive, can miss small or intermittent leaks, and cause new spinal leaks, the Bern Score is the new standard for diagnosing spinal leaks.
AI integrated into the Bern Score’s framework could help identify additional subtle signs of low CSF volume that a radiologist might miss, like pituitary enlargement and brain sagging. Through continuous learning, artificial intelligence can help improve the Bern Score’s diagnostic precision.
Human bias in radiology and its impacts on spinal leak diagnoses
Bias has been a critical issue that has contributed to many cases of spinal leaks being misdiagnosed.
For instance, radiologists often disregard signs of spinal leaks in women, like enlarged pituitary glands, labeling this engorgement as due to typical hormonal fluctuation in women despite this sign being a known symptom of CSF hypovolemia. A radiologist who specialized in MRIs of the brain for spinal leak evaluations once stated that pituitary size fluctuations in women are irrelevant, hinting at a pattern of missed diagnoses in his practice. Additionally, many radiologists ignore or underestimate tonsillar descent in patients.
Such biases are especially detrimental to women whose symptoms and pain are more likely to be dismissed as hormonal, exaggerated, or attributed to stress.
Another issue in diagnosing spinal leaks is having a single physician assess MRIs despite the well-known value of having multiple assessors for any type of evaluation. For instance, educational assessments typically require multiple assessors to ensure the effectiveness of tests. The signs of spinal leaks are subtle, and secondary checks would increase the accuracy of diagnoses.
Reducing human diagnostic bias with artificial intelligence
AI solves human bias by providing a data-driven approach that can be tested and calibrated for accuracy. Unlike humans, artificial intelligence can be trained to not suffer from gender or age bias. Also, its ability to analyze images isn’t affected by fatigue.
Bias in the training data AI learns from can be evaluated and tested regularly to ensure they deliver fair assessments for diverse patient populations.
Artificial intelligence can be an unbiased second opinion to help spot symptoms a radiologist has overlooked. AI has already started showing promise in challenging the interpretation of human radiologists, who dismissed critical signs of spinal leaks like 4 mm tonsillar descent.
Benefits of AI as an additional diagnostic assessor
AI reviewing imaging scans alongside radiologists can help catch subtle signs of spinal leaks that might go unnoticed. For example, AI can identify symptoms like low CSF volume, enlarged pituitary glands, or brain sagging when evaluating a patient’s brain MRIs, ensuring these critical signs aren’t dismissed due to bias.
Overcoming gender bias in diagnosing spinal leaks with AI
Women are disproportionately impacted by diagnostic bias in detecting spinal leaks. Symptoms like hormonal imbalances and chronic headaches are often minimized by radiologists who attribute them to normal female hormone fluctuations rather than signs of spinal leaks.
AI can help to sidestep these biases by focusing only on the MRI data without preconceived notions about typical symptom presentation or gender.
The future of AI in reducing diagnostic bias in spinal leak detection
The potential to improve diagnostic accuracy and reduce bias becomes more evident as artificial intelligence evolves. AI’s ability to counter human bias holds significant promise for improving patient care, particularly for those suffering from often misdiagnosed conditions like spinal leaks.
The advancement of artificial intelligence as a diagnostic tool is especially valuable to anesthesiologists looking to start private pain management clinics specializing in spinal leak care, where accurate CSF leak diagnosis and long-term comprehensive treatments can set them apart in an underserved market valued at almost a trillion dollars.
