Transformer versus traditional natural language processing: how much data is enough for automated radiology report classification?

OBJECTIVES: Current state-of-the-art natural language processing (NLP) techniques use transformer deep-learning architectures, which depend on large training datasets. We hypothesized that traditional NLP techniques may outperform transformers for smaller radiology report datasets. METHODS: We compared the performance of BioBERT, a deep-learning-based transformer model pre-trained on biomedical text, and three traditional machine-learning models (gradient boosted tree, random forest, and logistic regression) on seven classification tasks given free-text radiology reports. Tasks included detection of appendicitis, diverticulitis, bowel obstruction, and enteritis/colitis on abdomen/pelvis CT reports, ischemic infarct on brain CT/MRI reports, and medial and lateral meniscus tears on knee MRI reports (7,204 total annotated reports). The performance of NLP models on held-out test sets was compared after training using the full training set, and 2.5%, 10%, 25%, 50%, and 75% random subsets of the training data. RESULTS: In all tested classification tasks, BioBERT performed poorly at smaller training sample sizes compared to non-deep-learning NLP models. Specifically, BioBERT required training on approximately 1,000 reports to perform similarly or better than non-deep-learning models. At around 1,250 to 1,500 training samples, the testing performance for all models began to plateau, where additional training data yielded minimal performance gain. CONCLUSIONS: With larger sample sizes, transformer NLP models achieved superior performance in radiology report binary classification tasks. However, with smaller sizes (<1000) and more imbalanced training data, traditional NLP techniques performed better. ADVANCES IN KNOWLEDGE: Our benchmarks can help guide clinical NLP researchers in selecting machine-learning models according to their dataset characteristics.

Improving the patient experience during musculoskeletal interventional procedures.

In the patient-centered practice of musculoskeletal interventional radiology, pre-procedure, intra-procedure, and post-procedure factors impact the patient's overall experience. Opportunities to improve the patient experience begin before the patient procedure starts, with smooth scheduling of the appointment, a positive office environment, and appropriate communication by the musculoskeletal interventional radiology staff before the procedure, including clear expectations regarding wait times. The initial conversation between the radiologist and the patient, including the informed consent process, is a crucial opportunity for creating rapport with the patient and generating trust in the radiologist. Being able to interpret a patient's verbal and non-verbal cues during this conversation can reduce anxiety and enhance relaxation. During the procedure, there are interventions which can decrease anxiety and reduce the perception of pain by the patient. These include psychological interventions such as self-induced hypnosis, communication-based interventions such as empathic communication to mold a patient's interpretation of anxiety and painful stimuli, and the use of ancillary tools such as music or video stimulation. Finally, clear post-procedure expectations regarding post-procedure pain and its management and assurances of access to the musculoskeletal interventional radiology team can facilitate an uncomplicated post-procedure recovery.

Discitis-osteomyelitis: optimizing results of percutaneous sampling.

Vertebral discitis-osteomyelitis is an infection of the spine that involves the intervertebral disc and the adjacent vertebral body but may also extend into the paraspinal and epidural soft tissues. If blood cultures and other culture data fail to identify a causative microorganism, percutaneous sampling is indicated to help guide targeted antimicrobial therapy. Despite limited supporting evidence, withholding antimicrobial therapy for up to 2 weeks is recommended to maximize microbiological yield, although literature supporting this recommendation is limited. During the procedure, technical factors that may improve yield include targeting of paraspinal fluid collections or soft tissue abnormalities for sampling, acquiring multiple core samples if possible, and use of larger gauge needles when available. Repeat sampling may be indicated if initial percutaneous biopsy is negative but should be performed no sooner than 72 h after the initial percutaneous biopsy to ensure adequate time for culture results to return.

Automated Radiology-Arthroscopy Correlation of Knee Meniscal Tears Using Natural Language Processing Algorithms.

RATIONALE AND OBJECTIVES: Train and apply natural language processing (NLP) algorithms for automated radiology-arthroscopy correlation of meniscal tears. MATERIALS AND METHODS: In this retrospective single-institution study, we trained supervised machine learning models (logistic regression, support vector machine, and random forest) to detect medial or lateral meniscus tears on free-text MRI reports. We trained and evaluated model performances with cross-validation using 3593 manually annotated knee MRI reports. To assess radiology-arthroscopy correlation, we then randomly partitioned this dataset 80:20 for training and testing, where 108 test set MRIs were followed by knee arthroscopy within 1 year. These free-text arthroscopy reports were also manually annotated. The NLP algorithms trained on the knee MRI training dataset were then evaluated on the MRI and arthroscopy report test datasets. We assessed radiology-arthroscopy agreement using the ensembled NLP-extracted findings versus manually annotated findings. RESULTS: The NLP models showed high cross-validation performance for meniscal tear detection on knee MRI reports (medial meniscus F1 scores 0.93-0.94, lateral meniscus F1 scores 0.86-0.88). When these algorithms were evaluated on arthroscopy reports, despite never training on arthroscopy reports, performance was similar, though higher with model ensembling (medial meniscus F1 score 0.97, lateral meniscus F1 score 0.99). However, ensembling did not improve performance on knee MRI reports. In the radiology-arthroscopy test set, the ensembled NLP models were able to detect mismatches between MRI and arthroscopy reports with sensitivity 79% and specificity 87%. CONCLUSION: Radiology-arthroscopy correlation can be automated for knee meniscal tears using NLP algorithms, which shows promise for education and quality improvement.

Observer agreement in the choice of lumbar spine injection for pain management.

BACKGROUND AND PURPOSE: Lumbar spine MRI can help guide the choice of corticosteroid injection in pain management. We investigated whether patient-reported symptom information from a questionnaire could improve agreement in the choice of type, level, and side of injection. MATERIALS AND METHODS: In this prospective observational study, 120 patients (median age 64, 70 men) were recruited from patients referred for pain management. After informed consent, they completed electronic questionnaires that obtained symptom information for later use during MRI reviews. In 3 research arms, 6 radiologists chose injections that would ideally deliver corticosteroid to the presumed sources of pain in (1) MRI studies reviewed with symptom information from questionnaires, (2) MRI studies reviewed without symptom information, and (3) MRI reports. Blinded to questionnaire results, radiologists providing clinical care and interviewing patients chose ideal therapeutic injections to establish reference standards. Injections were categorized by type, level, and side and compared using percent agreement and kappa statistics. Interreading agreement was analyzed. RESULTS: Compared to the reference standard, kappa agreements for injection types, levels, and sides were almost perfect when MRIs were reviewed knowing symptoms (0.85-0.93), fair without symptoms (0.23-0.35) (all P < .001) and fair in MRI reports (0.24-0.36) (all P < .001). Interreading kappa agreements were almost perfect knowing symptoms (0.82-0.90), but only moderate without symptoms (0.42-0.49) (all P < .001). CONCLUSIONS: Radiologists reviewing lumbar spine MRI converged on the type, level, and side of ideal therapeutic injection whether they obtained symptom information from direct patient interview or electronic questionnaire. Observer agreement was significantly lower without symptom information.

Cryoablation for Palliation of Painful Bone Metastases: The MOTION Multicenter Study.

Purpose: To assess the clinical effectiveness of cryoablation for palliation of painful bone metastases. Materials and Methods: MOTION (Multicenter Study of Cryoablation for Palliation of Painful Bone Metastases) (ClinicalTrials.gov NCT02511678) was a multicenter, prospective, single-arm study of adults with metastatic bone disease who were not candidates for or had not benefited from standard therapy, that took place from February 2016 to March 2018. At baseline, participants rated their pain using the Brief Pain Inventory-Short Form (reference range from 0 to 10 points); those with moderate to severe pain, who had at least one metastatic candidate tumor for ablation, were included. The primary effectiveness endpoint was change in pain score from baseline to week 8. Participants were followed for 24 weeks after treatment. Statistical analyses included descriptive statistics and logistic regression to evaluate changes in pain score over the postprocedure follow-up period. Results: A total of 66 participants (mean age, 60.8 years ± 14.3 [standard deviation]; 35 [53.0%] men) were enrolled and received cryoablation; 65 completed follow-up. Mean change in pain score from baseline to week 8 was -2.61 points (95% CI: -3.45, -1.78). Mean pain scores improved by 2 points at week 1 and reached clinically meaningful levels (more than a 2-point decrease) after week 8; scores continued to improve throughout follow-up. Quality of life improved, opioid doses were stabilized, and functional status was maintained over 6 months. Serious adverse events occurred in three participants. Conclusion: Cryoablation of metastatic bone tumors provided rapid and durable pain palliation, improved quality of life, and offered an alternative to opioids for pain control.Keywords: Ablation Techniques, Metastases, Pain Management, Radiation Therapy/OncologySupplemental material is available for this article.© RSNA, 2021.

Impact of Patient-Reported Symptom Information on Agreement in the MRI Diagnosis of Presumptive Lumbar Spine Pain Generator.

BACKGROUND. Lumbar spine MRI is associated with a high prevalence of interpretive errors by radiologists. Treating physicians can obtain symptom information, correlate symptoms with MRI findings, and distinguish presumptive pain generators from incidental abnormalities. OBJECTIVE. The purpose of this study was to capture symptom information using a patient questionnaire, review lumbar spine MRI examinations with and without symptom information, diagnose pain generators, and compare MRI diagnoses with clinical reference diagnoses. METHODS. In this prospective study, 120 participants (70 men and 50 women; median age, 64 years; interquartile range, 49.5-74 years) were recruited from patients referred for lumbar spine injections between February and June 2019. Participants completed electronic questionnaires regarding their symptoms before receiving the injections. For three research arms, six radiologists diagnosed pain generators in MRI studies reviewed with symptom information from questionnaires, MRI studies reviewed without symptom information, and MRI reports. Interreading agreement was analyzed. Blinded to the questionnaire results, the radiologists who performed injections obtained patient histories, correlated symptoms with MRI findings, and diagnosed presumptive pain generators. These diagnoses served as clinical reference standards. Pain generators were categorized by type, level, and side and were compared using kappa statistics. Diagnostic certainty was recorded using numeric values (0-100) and was compared using Wilcoxon rank-sum test RESULTS. When compared with the reference standard, agreement for the type, level, and side of pain generator was almost perfect in MRI examinations reviewed with symptom information (κ = 0.82-0.90), fair to moderate in MRI examinations reviewed without symptom information (κ = 0.28-0.51) (all p < .001), and fair to moderate in MRI reports (κ = 0.27-0.45) (all p < .001). Interreading agreement was almost perfect when MRI examinations were reviewed with symptom information (κ = 0.82-0.90) but was only moderate without symptom information (κ = 0.42-0.56) (all p < .001). Diagnostic certainty levels were highest for radiologists performing injections (mean [± SD], 90.0 ± 9.9) and were significantly higher for MRI review with symptom information versus without symptom information (means for reading 1, 84.6 ± 13.1 vs 62.9 ± 20.7; p < .001). CONCLUSION. In lumbar spine MRI, presumptive pain generators diagnosed using symptom information from electronic questionnaires showed almost perfect agreement with pain generators diagnosed using symptom information from direct patient interviews. CLINICAL IMPACT. Patient-reported symptom information from a brief questionnaire can be correlated with MRI findings to distinguish presumptive pain generators from incidental abnormalities.

Siamese neural networks for continuous disease severity evaluation and change detection in medical imaging.

Using medical images to evaluate disease severity and change over time is a routine and important task in clinical decision making. Grading systems are often used, but are unreliable as domain experts disagree on disease severity category thresholds. These discrete categories also do not reflect the underlying continuous spectrum of disease severity. To address these issues, we developed a convolutional Siamese neural network approach to evaluate disease severity at single time points and change between longitudinal patient visits on a continuous spectrum. We demonstrate this in two medical imaging domains: retinopathy of prematurity (ROP) in retinal photographs and osteoarthritis in knee radiographs. Our patient cohorts consist of 4861 images from 870 patients in the Imaging and Informatics in Retinopathy of Prematurity (i-ROP) cohort study and 10,012 images from 3021 patients in the Multicenter Osteoarthritis Study (MOST), both of which feature longitudinal imaging data. Multiple expert clinician raters ranked 100 retinal images and 100 knee radiographs from excluded test sets for severity of ROP and osteoarthritis, respectively. The Siamese neural network output for each image in comparison to a pool of normal reference images correlates with disease severity rank (ρ = 0.87 for ROP and ρ = 0.89 for osteoarthritis), both within and between the clinical grading categories. Thus, this output can represent the continuous spectrum of disease severity at any single time point. The difference in these outputs can be used to show change over time. Alternatively, paired images from the same patient at two time points can be directly compared using the Siamese neural network, resulting in an additional continuous measure of change between images. Importantly, our approach does not require manual localization of the pathology of interest and requires only a binary label for training (same versus different). The location of disease and site of change detected by the algorithm can be visualized using an occlusion sensitivity map-based approach. For a longitudinal binary change detection task, our Siamese neural networks achieve test set receiving operator characteristic area under the curves (AUCs) of up to 0.90 in evaluating ROP or knee osteoarthritis change, depending on the change detection strategy. The overall performance on this binary task is similar compared to a conventional convolutional deep-neural network trained for multi-class classification. Our results demonstrate that convolutional Siamese neural networks can be a powerful tool for evaluating the continuous spectrum of disease severity and change in medical imaging.

Percutaneous Cryoablation: Safety and Efficacy for Pain Palliation of Metastases to Pleura and Chest Wall.

PURPOSE: To assess safety and efficacy of percutaneous cryoablation for pain palliation of metastases to pleura and chest wall. MATERIALS AND METHODS: This retrospective single-center cohort study included 22 patients (27% female, mean age 63 y ± 11.4) who underwent 25 cryoablation procedures for pain palliation of 39 symptomatic metastases measuring 5.1 cm ± 1.9 (range, 2.0-8.0 cm) in pleura and chest wall between June 2012 and December 2017. Pain intensity was assessed using a numerical scale (0-10 points). Statistical tests t test, χ2, and Wilcoxon signed rank were performed. RESULTS: Patients were followed for a median of 4.1 months (interquartile range [IQR], 2.3-10.1; range, 0.1-36.7 mo) before death or loss to follow-up. Following cryoablation, pain intensity decreased significantly by a median of 4.5 points (IQR, 2.8-6; range, 0-10 points; P = .0002 points, Wilcoxon signed rank). Pain relief of at least 3 points was documented following 18 of 20 procedures. Pain relief occurred within a median of 1 day following cryoablation (IQR, 1-2; range, 1-4 d) and lasted for a median of 5 weeks (IQR, 3-17; range, 1-34 wk). Systemic opioid requirements decreased in 11 of 22 patients (50%) by an average of 56% ± 34. Difference in morphine milligram equivalents was not significant (P = .73, Wilcoxon signed rank). No procedure-related complications occurred despite previous radiation of 7 tumors. Of 25 procedures, 22 (88%) were performed on an outpatient basis. CONCLUSIONS: Percutaneous cryoablation for metastases to pleura and chest wall can safely provide significant pain relief within days following a single session.

The effect of music on pain and subjective experience in image-guided musculoskeletal corticosteroid injections: a randomized controlled trial.

PURPOSE: To investigate the role of music on subjects undergoing routine image-guided musculoskeletal corticosteroid injections and its effect on post-procedure pain and subjective overall experience. MATERIALS AND METHODS: This prospective study was IRB-approved and HIPAA-compliant. A total of 126 subjects referred for outpatient image-guided musculoskeletal corticosteroid injections were enrolled in the study and randomized into a music offered group ((+)MO) and a no music offered group ((-)MO). (+)MO subjects were given the opportunity to listen to music during their corticosteroid injection. All subjects were then given an anonymous survey on which they recorded their pre-procedural and post-procedural pain on a scale from 0 to 9 and rated their overall experience and how likely they were to recommend our department for musculoskeletal procedures on scales from 1 to 5. RESULTS: (+)MO subjects had significantly lower post-procedural pain (p = 0.013) and significantly greater decrease in pain (p = 0.031) compared to (-)MO subjects. Among the (+)MO subjects, there was no statistically significant difference in post-procedure pain (p = 0.34) or change in pain (p = 0.62) if music was accepted or declined. However, subjects who listened to music did have lower post-procedural pain compared to those who did not listen to music (p = 0.012), although the differences in the decrease of pain between the two groups did not quite reach statistical significance (p = 0.062). CONCLUSIONS: Playing music during image-guided musculoskeletal corticosteroid injections may reduce patients' post-procedure pain. Offering patients some measure of control over their procedure may be a factor that contributes to decreased post-procedure pain as well.

Ultrasound-guided injection for the diagnosis and treatment of posteromedial knee friction syndrome.

OBJECTIVE: To describe an ultrasound guided injection technique for diagnosing and treating posteromedial knee friction syndrome, which occurs between the sartorius/gracilis tendons and medial femoral condyle (MFC). MATERIALS AND METHODS: Our study was IRB-approved and HIPAA-compliant. We identified patients via a retrospective review of medical records and MRI with posteromedial knee pain and isolated edema between MFC and sartorius/gracilis tendons and no evidence for meniscal tear, ruptured Baker's cyst or degenerative joint disease. Patients were referred for an ultrasound-guided procedure to inject anesthetic and corticosteroid at the site of edema. Procedures were evaluated for technical success, which was defined as satisfactory identification of the injection site and adequate delivery of medication. Follow-up was available up to 8 weeks after the procedure to determine the response and any potential complications. RESULTS: Fourteen subjects with MRI and symptoms of posteromedial knee friction syndrome underwent 14 injections. Technical success was achieved in all procedures, with no complications. At 8 weeks' follow-up, 92% of patients had symptom improvement. VAS before and 8 weeks after the procedure changed from 5.2 ± 2.7 to 0.9 ± 2.1 (p = 0.0002), respectively. CONCLUSION: Ultrasound-guided injection of edema between the MFC and sartorius/gracilis tendons supports the diagnosis of a posteromedial knee friction syndrome and successfully treats its associated symptoms.

Diagnostic yield of repeat CT-guided biopsy for suspected infectious spondylodiscitis.

OBJECTIVE: The purpose of this study was to examine the yield of repeat CT-guided biopsy in patients with suspected infectious spondylodiscitis following an initial biopsy with negative microbiology, and to identify factors associated with successful pathogen isolation. MATERIALS AND METHODS: In this retrospective study, 21 patients (12 men, 9 women; mean age, 52, range, 12-84) were identified with clinically and radiologically suspected infectious spondylodiscitis who underwent repeat biopsy following negative cultures from an initial biopsy. The microbe yield as well as demographic, clinical, and laboratory findings were reviewed and statistical analysis was performed. RESULTS: Repeat CT-guided biopsy isolated a causative microbe in 3/21 patients (14.3%). Younger age (p = 0.021) was significantly associated with successful microbe isolation. All three cases of successful microbe isolation occurred in patients not exposed to antibiotics (3/9 patients) whereas no successful microbe isolation occurred in patients who received antibiotics (0/12 patients); however, this difference did not reach statistical significance (p = 0.062). Gender, duration of symptoms, white blood cell count, biopsy interval, and biopsy site were not significantly associated with microbe isolation. CONCLUSIONS: Overall microbiologic yield of repeat CT-guided biopsy for patients with suspected infectious spondylodiscitis was low at 14.3%; however, a higher yield was identified in patients who were younger in age and not exposed to pre-biopsy antibiotics.

The Effect of Computer-Assisted Reporting on Interreader Variability of Lumbar Spine MRI Degenerative Findings: Five Readers With 30 Disc Levels.

PURPOSE: The aim of this study was to assess differences in interreader variability among radiologists after the implementation of a computer-assisted reporting (CAR) tool for the interpretation of degenerative disc disease on lumbar spine MRI. METHODS: Thirty lumbar spine MRI examinations were selected from the radiology database. Five fellowship-trained musculoskeletal radiologists evaluated each L4-L5 disc in a blinded fashion and reported the findings using a traditional free dictation approach. One month later, they reinterpreted the same discs using a web browser-based CAR tool in the same blinded fashion. The degrees of central canal stenosis, neural foraminal stenosis, and facet joint osteoarthritis; presence or absence of lateral recess stenosis; types of disc bulge or herniation; and herniation location using both methods were recorded. Percentage disagreement among the radiologists for each variable was calculated and compared using the Wilcoxon signed rank test. RESULTS: There was a statistically significant decrease among the five radiologists in percentage disagreement for neural foraminal stenosis (46% versus 35%, P = .0146) and facet joint osteoarthritis (45% and 22%, P < .0001) for reports created by free dictation compared with those created using the CAR tool. There was no statistically significant difference in interreader variability for the assessment of central canal stenosis, lateral recess effacement, disc herniation, disc bulge, or herniation location. CONCLUSIONS: Implementation of a CAR tool for the interpretation of degenerative changes on lumbar spine MRI decreases interreader variability in the assessment of neural foraminal stenosis and facet joint osteoarthritis.

Percutaneous CT-guided sacroiliac joint sampling for infection: aspiration, biopsy, and technique.

OBJECTIVE: To evaluate methods of CT-guided sacroiliac joint sampling in patients with suspected infection. MATERIALS AND METHODS: All CT-guided sacroiliac joint sampling procedures for suspected infection were reviewed for sampling type (aspiration, lavage aspiration, biopsy), microbiology results, and clinical and imaging follow-up. The primary gold standard was anatomic pathology. If pathology was not available, then positive blood culture with the same organism as SIJ sampling, imaging and clinical follow-up, or clinical follow-up only were used. Anterior and posterior joint distention was evaluated by MRI within 7 days of the procedure. RESULTS: A total of 34 patients (age 39 ± 20 (range, 6-75) years; 21 F, 13 M) were included. Aspiration samples only were obtained in 13/34 (38%) cases, biopsy samples only in 9/34 (26%) cases, and both samples in 12/34 (35%) cases. There was an overall 54% sensitivity and 86% specificity. For the aspiration samples, sensitivity and specificity were 60 and 81%, respectively, compared to 45 and 90% for the biopsy samples. In cases with both samples, biopsy did not add additional microbial information. Seventeen (17/34, 50%) patients had an MRI. The anterior joint was more distended than the posterior joint in 15/17 (88%) of patients, and this difference was significant (P = 0.0003). All of these 17 patients had an attempted aspiration by a posterior approach; 6/17 (35%) resulted in a successful aspiration. CONCLUSIONS: Aspiration of the sacroiliac joint has a higher sensitivity than biopsy and should always be attempted first. MRI may be helpful for procedure planning.

Variations of anatomy on MRI of the first extensor compartment of the wrist and association with DeQuervain tenosynovitis.

OBJECTIVE: To study anatomical variations on MRI of the first extensor compartment of the wrist in DeQuervain tenosynovitis (DQT). MATERIALS AND METHODS: A retrospective search for DQT patients yielded 47 subjects (51 ± 15 years, 36 female, 11 male). The age-matched control group (normal first extensor compartment) was 49 ± 15 years (29 female, 18 male). Two independent readers reviewed: the number of abductor pollicis longus (APL) tendon slips, tendon sheath septations (compartmentalization), and APL and EPB cross-sectional area (CSA) at the radial styloid. A tendon slip was defined as a discrete structure for ≥5 contiguous slices with its own insertion. RESULTS: The distribution of APL tendon slips was different for the DQT and control groups (Reader 1/Reader 2: P = 0.0001 and 0.001). The most common arrangement for both groups was two APL tendon slips. One tendon slip was less common (P = 0.03 and 0.1) and compartmentalization was more common (P = 0.003; < 0.0001) for the DQT group than the control group. There was no difference in tendon slip insertions on one or multiple bones (P = 0.1; 0.7). APL and EPB compartment CSAs were also higher for the DQT group (combined first extensor compartment area: 21.3 ± 7.6 mm2; 21.0 ± 7.1) than the control group (17.2 ± 3.8; 17.1 ± 3.9) (P = 0.002; 0.002). CONCLUSION: We found a statistically significantly increased proportion of supernumerary tendon slips and compartmentalization of the first extensor compartment in patients with DQT and greater CSA of the first extensor compartment at the radial styloid, consistent with previous anatomical, surgical, and ultrasound studies.

Successful CT guided cryoablation of phosphaturic mesenchymal tumor in the soft tissues causing tumor-induced osteomalacia: a case report.

We report the case of a 25-year-old male patient with tumor-induced osteomalacia from a 1.8-cm phosphaturic mesenchymal tumor in the right distal thigh, who was treated at our institution with a single session of CT-guided cryoablation in December 2015, which resulted in biochemical and clinical resolution. We present the clinical history, physical examination, biochemistry, functional imaging, anatomic characterization, and follow-up for clinical outcome. The response to treatment was documented in terms of normalization of serum fibroblastic growth factor 23 (FGF23) and phosphorous levels, symptomatic improvement, as well as normalization of bone mineralization on femur radiographs 3 months after the procedure. Although the first-line treatment for phosphaturic mesenchymal tumor-induced osteomalacia is wide surgical excision, CT-guided cryoablation widens the array of treatment options, especially in those patients who decline surgery or who are otherwise poor surgical candidates.

Treatment of aneurysmal bone cysts by percutaneous CT-guided injection of calcitonin and steroid.

OBJECTIVES: To determine the efficacy and safety of percutaneous calcitonin and steroid injection in the treatment of aneurysmal bone cysts (ABCs). MATERIALS AND METHODS: Our study was IRB-approved and HIPAA-compliant. We reviewed pre- and post-procedural imaging studies and medical records of all CT-guided percutaneous injections of ABCs with calcitonin and steroid performed at our institution between 2003 and 2015. RESULTS: Treatment success based on imaging was categorized as substantial (51-100 %), partial (1-50 %), or none (0 %) by comparing radiographs of the lesion before and after treatment. Our study group comprised 9 patients (7 female, 2 male; mean age 19 ± 5 (range 12-25) years). ABCs were located in the pubis (n = 3), femur (n = 2), and humerus/scapula/ilium/sacrum (n = 1 for each). One patient did not have any clinical or imaging follow-up. For the other 8 patients, clinical and imaging follow-up ranged from 1 to 93 months (mean 16 ± 29 months). One patient had two injections, and 1 patient had three injections. Six out of eight patients (75 %) had complete symptomatic relief and 2 patients (25 %) had partial symptomatic relief after initial injection. Imaging follow-up revealed substantial imaging response in 4 out of 8 patients (50 %). There was a partial imaging response in 2 patients (25 %) and no imaging response in 2 out of 8 patients (25 %), and all 4 of these patients had local recurrence. There were no complications. CONCLUSION: Percutaneous CT-guided injection of ABCs with calcitonin and steroid is a safe and effective treatment. Lack of imaging response may necessitate more aggressive treatment to minimize local recurrence.