Deep learning to detect anterior cruciate ligament tear on knee MRI: multi-continental external validation

OBJECTIVES

To develop a deep-learning algorithm for anterior cruciate ligament (ACL) tear detection and to compare its accuracy using two external datasets.

METHODS

A database of 19,765 knee MRI scans (17,738 patients) issued from different manufacturers and magnetic fields was used to build a deep learning-based ACL tear detector. Fifteen percent showed partial or complete ACL rupture. Coronal and sagittal fat-suppressed proton density or T2-weighted sequences were used. A Natural Language Processing algorithm was used to automatically label reports associated with each MRI exam. We compared the accuracy of our model on two publicly available external datasets: MRNet, Bien et al, USA (PLoS Med 15:e1002699, 2018); and KneeMRI, Stajduhar et al, Croatia (Comput Methods Prog Biomed 140:151-164, 2017). Receptor operating characteristics (ROC) curves, area under the curve (AUC), sensitivity, specificity, and accuracy were used to evaluate our model.

RESULTS

Our neural networks achieved an AUC value of 0.939 for detection of ACL tears, with a sensitivity of 87% (0.875) and a specificity of 91% (0.908). After retraining our model on Bien dataset and Stajduhar dataset, our algorithm achieved AUC of 0.962 (95% CI 0.930-0.988) and 0.922 (95% CI 0.875, 0.962) respectively. Sensitivity, specificity, and accuracy were respectively 85% (95% CI 75-94%, 0.852), 89% (95% CI 82-97%, 0.894), 0.875 (95% CI 0.817-0.933) for Bien dataset, and 68% (95% CI 54-81%, 0.681), 93% (95% CI 89-97%, 0.934), and 0.870 (95% CI 0.821-0.913) for Stajduhar dataset.

CONCLUSION

Our algorithm showed high performance in the detection of ACL tears with AUC on two external datasets, demonstrating its generalizability on different manufacturers and populations. This study shows the performance of an algorithm for detecting anterior cruciate ligament tears with an external validation on populations from countries and continents different from the study population.

KEY POINTS

• An algorithm for detecting anterior cruciate ligament ruptures was built from a large dataset of nearly 20,000 MRI with AUC values of 0.939, sensitivity of 87%, and specificity of 91%. • This algorithm was tested on two external populations from different other countries: a dataset from an American population and a dataset from a Croatian population. Performance remains high on these two external validation populations (AUC of 0.962 and 0.922 respectively).

Superficial myxofibrosarcoma: assessment of recurrence risk according to the surgical margin following resection. A series of 21 patients

INTRODUCTION

Superficial myxofibrosarcomas are malignant connective tissue tumors, whose very frequent recurrence influences the local and vital prognosis. Even when resection seems to be macroscopically complete it is very often microscopically contaminated. The aim of this study was to evaluate recurrence in relation to the surgical margins and to compare, when possible, tumor size, evaluated clinically and macroscopically by the pathologist.

MATERIALS AND METHODS

This was a single center study of 21 patients, mean age 67 years old, treated for superficial myxofibrosarcoma. The number, date and location of recurrence were collected for each patient. A clinical and pathological measurement was made of the longest axis of the tumor in each case of recurrence.

RESULTS

Fifty-seven percent of patients presented with recurrent tumors. The mean number of recurrences was 1.4 per patient (1-8). The surgical margins were wide in four cases, marginal in two cases and incomplete/intralesional in 15 other patients with a rate of recurrence of 25, 50 and 67% respectively. The size evaluated during the preoperative clinical examination (14 cases) was underestimated by a mean 2.4 cm compared to the macroscopic pathology assessment. The preoperative size on MRI (5 cases) was also underestimated by a mean 1.3 cm.

CONCLUSION

Superficial myxofibrosarcomas are tumors that are difficult to resect completely because they are infiltrative, a feature that is often underestimated before surgery. Surgical treatment of this entity requires a much larger surgical margin than that suggested by the preoperative clinical and MRI evaluations. In case of incomplete resection, revision scar surgery should systematically be performed.

LEVEL OF EVIDENCE

Level IV. Retrospective study.

Primary or recurring extra-abdominal desmoid fibromatosis: assessment of treatment by observation only

INTRODUCTION

Extra-abdominal desmoid fibromatosis (EADF) is a benign tumoral condition, classically managed by more or less radical and sometimes mutilating excision. This treatment strategy is associated with a recurrence rate of nearly 50% according to various reports.

HYPOTHESIS

EADF may show spontaneous stabilization over time.

METHODS

A retrospective series of 26 cases of EADF managed by simple observation was studied to assess spontaneous favorable evolution and identify possible factors impacting evolution. Eleven cases were of primary EADF with no treatment or surgery, and 15 of recurrence after surgery with no adjuvant treatment. MRI was the reference examination during follow-up.

RESULTS

Twenty-four cases showed stabilization at a median 14 months; there were no cases of renewed evolution after stabilization. One primary tumor showed spontaneous regression, and one recurrence still showed evolution at end of follow-up (23 months). The sole factor impacting potential for evolution was prior surgery. No radiologic or pathologic criteria of evolution emerged from analysis.

DISCUSSION

The present series, one of the largest dedicated to EADF managed by observation, confirmed recent literature findings: a conservative "wait-and-see" attitude is reasonable and should be considered when large-scale resection would entail significant functional or esthetic impairment.

LEVEL OF EVIDENCE

Level IV, retrospective study.

[Dedifferentiated chondrosarcoma: radiologic-pathologic correlation]

Dedifferentiated chondrosarcomas are highly malignant tumors characterized by conventional low-grade chondrosarcoma with abrupt transition to foci that have dedifferentiated into a higher-grade noncartilaginous more aggressive sarcoma. The dedifferentiated component, an osteosarcoma or fibrosarcoma, determines the prognosis. Its identification is key for management. A diagnosis of dedifferentiated chondrosarcoma should be suggested by the presence of "tumoral dimorphism" with cartilaginous component and aggressive lytic component invading adjacent soft tissues.

Role of imaging in spine, hand, and wrist osteoarthritis

Osteoarthritis (OA) of the wrist is mainly secondary to traumatic ligamentous or bone injuries. Involvement of the radiocarpal joint occurs early on in the disease, whereas the mediocarpal joint is involved at a later stage. Metabolic diseases may also involve the wrist and affect specific joints such as the scapho-trapezio-trapezoid joint. Although OA of the wrist is routinely diagnosed on plain films, a thorough assessment of cartilage injuries on computed tomographic arthrography, magnetic resonance imaging (MRI), or MR arthrography remains necessary before any surgical procedure. OA of the fingers is frequently encountered in postmenopausal women. Distal interphalangeal joints and trapezio-metacarpal joint are the most frequently involved joints. Whereas the clinical diagnosis of OA of the wrist and hand is straightforward, the therapeutic management of symptomatic forms remains unclear, with no clear guidelines. OA of the spine is related to degenerative changes of the spine involving the disc space, vertebral endplates, the facet joints, or the supportive and surrounding soft tissues. The sequelae of disc degeneration are among the leading causes of functional incapacity in both sexes, and are a common source of chronic disability in the working years. Disc degeneration involves structural disruption and cell-mediated changes in composition. Radiography remains usually the first-line imaging method. MRI is ideally suited for delineating the presence, extent, and complications of degenerative spinal disease. Other imaging modalities such as computed tomography, dynamic radiography, myelography, and discography may provide complementary information in selected cases, especially before an imaging-guided percutaneous treatment or spinal surgery. The presence of degenerative changes on imaging examinations is by no means an indicator of symptoms, and there is a high prevalence of lesions in asymptomatic individuals. This article focuses on imaging of OA of the wrist and hand, as well as lumbar spine OA, with an emphasis on current MRI grading systems available for the assessment of discovertebral lesions.

Hip arthroplasty. Part 2: normal and abnormal radiographic findings

This review addresses the normal and abnormal radiographic findings that can be encountered during the follow-up of patients with total hip arthroplasty (THA). The relative significance of different patterns of radiolucency, bone sclerosis, and component position is discussed. The normal or pathological significance of these findings is correlated with design, surface, and fixation of the prosthetic components. It is essential to have a good knowledge of expected and unexpected radiological evolution according to the different types of prostheses. This paper emphasizes the importance of serial studies compared with early postoperative radiographs during follow-up in order to report accurately any sign of prosthetic failure and trigger prompt specialist referral. Basic technical guidelines and schedule recommendations for radiological follow-up are summarized.

Uterine Fibroid Vascularization and Clinical Relevance to Uterine Fibroid Embolization

Embolization has become a first-line treatment for symptomatic uterine fibroid tumors. Selective catheterization and embolization of both uterine arteries, which are the predominant source of blood flow to fibroid tumors in most cases, is the cornerstone of treatment. Although embolization for treatment of uterine fibroid tumors is widely accepted, great familiarity with the normal and variant pelvic arterial anatomy is needed to ensure the safety and success of the procedure. The uterine artery classically arises as a first or second branch of the anterior division of the internal iliac artery and is usually dilated in the presence of a uterine fibroid tumor. Angiography is used for comprehensive pretreatment assessment of the pelvic arterial anatomy; for noninvasive evaluation, Doppler ultrasonography, contrast material-enhanced magnetic resonance (MR) imaging, and MR angiography also may be used. After the uterine artery is identified, selective catheterization should be performed distal to its cervicovaginal branch. For targeted embolization of the perifibroid arterial plexus, injection of particles with diameters larger than 500 mum is generally recommended. Excessive embolization may injure normal myometrium, ovaries, or fallopian tubes and lead to uterine necrosis or infection or to ovarian failure. Incomplete treatment or additional blood supply to the tumor (eg, via an ovarian artery) may result in clinical failure. The common postembolization angiographic end point is occlusion of the uterine arterial branches to the fibroid tumor while antegrade flow is maintained in the main uterine artery.