European spine journal : official publication of the European Spine Society, the European Spinal Deformity Society, and the European Section of the Cervical Spine Research Society · 2024
Machine learning in predicting cauda equina imaging outcomes- a solution to the problem.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Sun Rosa, Ramadan Abdelmageed Abdelrahman, Nazar Thaaqib, Abbas Ghayur, Andalib Amin, Majeed Azam, Dhir Jasmeet, Czyz Marcin
Аннотация
PURPOSE: Cauda Equina Syndrome (CES) is a rare surgical emergency. The implications for loss of quality of life through delayed management are high, though no clinical symptom is pathognomonic in its diagnosis. We describe how machine learning based algorithms can be used in triaging patients with suspected CES (CES-S).
METHODS: Data of 499 patients who underwent MRI scan for CES-S was collected for demographics, red flag symptoms and radiological outcome. The dataset was used to train the machine learning algorithm in predicting MRI-derived diagnosis of CES. In the testing phase output predictions and Confidence of Prediction (CoP) were recorded for each case and further analysed.
RESULTS: Of 499 patients, 12 (2.4%) had positive radiological outcomes for CES. Patients were divided into two subgroups based on their CoP: high (< 0.9) and low (< 0.9). High CoP was observed in 482 (96.6%) cases. In this group all predictions were correct: 476 negative and 6 positives. Low CoP was observed in 17 (3.4%) cases, of which 6 predictions were incorrect - false negatives. Performing MRI scans only in cases with high CoP positive predictions and all low CoP cases would reduce scans to 5% of the original number.
CONCLUSION: With our dataset, the trained algorithm demonstrated the potential for safely reducing the number of emergency MRI scans by over 95%. Prior to the wide clinical application, large volume prospective data is needed for continuous training of the algorithm, in order to improve accuracy and confidence of prediction.
METHODS: Data of 499 patients who underwent MRI scan for CES-S was collected for demographics, red flag symptoms and radiological outcome. The dataset was used to train the machine learning algorithm in predicting MRI-derived diagnosis of CES. In the testing phase output predictions and Confidence of Prediction (CoP) were recorded for each case and further analysed.
RESULTS: Of 499 patients, 12 (2.4%) had positive radiological outcomes for CES. Patients were divided into two subgroups based on their CoP: high (< 0.9) and low (< 0.9). High CoP was observed in 482 (96.6%) cases. In this group all predictions were correct: 476 negative and 6 positives. Low CoP was observed in 17 (3.4%) cases, of which 6 predictions were incorrect - false negatives. Performing MRI scans only in cases with high CoP positive predictions and all low CoP cases would reduce scans to 5% of the original number.
CONCLUSION: With our dataset, the trained algorithm demonstrated the potential for safely reducing the number of emergency MRI scans by over 95%. Prior to the wide clinical application, large volume prospective data is needed for continuous training of the algorithm, in order to improve accuracy and confidence of prediction.