X-ray of mildly compressed hip fracture.

Science and technology
Tools
Typography

Researchers at the University of Turku and HUS Helsinki University Hospital are leading a revolution in medical diagnostics through the use of deep learning algorithms, significantly improving patient care by accurately identifying fractures, benign bone tumors, and necrosis in hand X-rays.

The introduction of AI in medical imaging, particularly in diagnosing conditions related to the hand and wrist, is showcased in three peer-reviewed international journal articles produced by HUS's hand surgery department.

These studies underline AI's potential in enhancing patient treatment by facilitating early and accurate diagnoses, crucial for preventing adverse outcomes.

Deep learning algorithms have been developed to identify wrist fractures, benign bone tumors, and bone necrosis in wrist X-rays. This AI technology offers valuable support in emergency departments and for physicians not specialized in interpreting hand or wrist X-rays, reducing the likelihood of misdiagnosis which can lead to unnecessary delays or complications in treatment.

AI Detects Fractures with High Accuracy

Distal radius fractures, constituting 20% of all emergency department fractures, require precise clinical examination and X-ray interpretation for diagnosis and treatment. Misinterpretation of X-rays is common, potentially leading to further investigations. AI can offer critical assistance in emergency settings, minimizing diagnostic errors. After training, an AI model developed by HUS's head of hand surgery, Dr. Jorma Ryhänen, and his research team accurately identified 97% of these fractures.

"The recognition of joint-disrupting comminuted fractures and decisions on further treatment can be challenging, sometimes resulting in delayed or improperly healed fractures. The AI model we've developed excellently identifies these fractures, potentially offering immediate recommendations for treatment, whether cast immobilization or surgery," states Dr. Ryhänen.

Supporting Emergency Departments in Accurate X-ray Interpretation

Enchondroma, a common benign bone tumor in the hand, can be challenging to diagnose in the presence of fractures, often going unnoticed in emergency situations. A team of hand surgeons led by Dr. Ryhänen developed an AI model capable of identifying enchondromas in 56 out of 62 test cases, proving to be a useful tool for emergency department physicians unfamiliar with hand-related ailments.

"The detection of enchondromas is an exciting first step towards automated diagnostics of bone tumors in the hand area, demonstrating that even with a limited number of images, it is possible to create effective diagnostic aids," shares hand surgeon Dr. Turkka Anttila, part of the research team.

AI's Precision Surpasses Human Eye

Another study focused on improving the diagnosis of lunate necrosis, a rare condition mainly affecting men aged 20–40. Early stages of necrosis are invisible to the human eye in X-rays, leading to delayed diagnoses and limited treatment options. The AI model developed by Dr. Ryhänen's team was able to identify necrosis in 28 out of 30 cases, outperforming a group of experienced specialists.

Deep learning algorithms are swiftly becoming indispensable tools in medical imaging, with the potential to transform medical classification and diagnostics in the coming decade. "It's clear that AI algorithms will significantly change patient diagnostics and treatment in surgical practice soon. AI can tirelessly and cost-effectively sift through large volumes of images, identifying abnormalities, risk factors, and complications, thus speeding up and improving treatment outcomes. High-level, interdisciplinary expertise and research are necessary to develop and validate reliable and generalizable models. Our research team has several promising ideas for further investigation," concludes Dr. Ryhänen.

HT

Partners