Truace
Subscribe

Deep Learning-Based Automated Computed Tomography Classification of Posterior Malleolar Fractures for a Standardized Diagnostic Framework

Despite advances in computed tomography (CT)-based classification systems for posterior malleolar fractures (PMFs), diagnostic consistency remains limited because these systems are still highly dependent on the observer's experience. This study aimed to develop a deep learning-based automated model capable of classifying PMFs according to the Haraguchi, Mason-Molloy, and Bartoníček-Rammelt clasification systems using CT scans, and to compare the classification performance of different convolutional neural networ…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-2019· cite & verify
Deep Learning-Based Automated Computed Tomography Classification of Posterior Malleolar Fractures for a Standardized Diagnostic Framework
Polio physical therapy by Photo Credit: Content Providers: CDC/Charles Farmer. Public domain

In brief

Researchers built deep learning models to sort posterior malleolar fractures into three established CT-based classification systems. They used preoperative scans from 480 adult patients, with three orthopedic surgeons' majority consensus as the reference standard, and compared VGG-16, ResNet-101, and DenseNet-121.

Existing CT classification systems depend heavily on the observer's experience, so consistent grading is hard to achieve. The authors conclude that automated classification is feasible and may support standardization, but the results come from retrospective data and do not show effects on patient care.

Main points

  1. The dataset covered 480 adult patients and 2398 CT slices, labeled by three orthopedic surgeons.
  2. Data were split 60% training, 20% validation, and 20% testing.
  3. VGG-16 and ResNet-101 performed well on the Haraguchi and Mason-Molloy systems.
  4. F1-scores were above 0.70 in most categories and above 0.80 in several.

The gain

Convolutional neural networks classified posterior malleolar fractures on CT with strong discrimination, suggesting automated grading could reduce observer-dependent variability.

The rundown

The study targeted the Haraguchi, Mason-Molloy, and Bartoníček-Rammelt systems, which remain dependent on observer experience despite advances in CT-based classification.

Anonymized DICOM images were fed to three convolutional architectures, and performance was measured by accuracy, sensitivity, precision, F1 score, and AUC.

VGG-16 reached particularly high AUC in Haraguchi types I-III, Mason-Molloy types IIB and III, and Bartoníček-Rammelt type IV, while ResNet-101 also exceeded 0.90 in most classes.

Sources

  1. Peer-reviewedJournal of Imaging Informatics in Medicine2026-10-13

The debate