Building RadiologyNET: an unsupervised approach to annotating a large-scale multimodal medical database.

BACKGROUND: The use of machine learning in medical diagnosis and treatment has grown significantly in recent years with the development of computer-aided diagnosis systems, often based on annotated medical radiology images. However, the lack of large annotated image datasets remains a major obstacle, as the annotation process is time-consuming and costly. This study aims to overcome this challenge by proposing an automated method for annotating a large database of medical radiology images based on their semantic similarity. RESULTS: An automated, unsupervised approach is used to create a large annotated dataset of medical radiology images originating from the Clinical Hospital Centre Rijeka, Croatia. The pipeline is built by data-mining three different types of medical data: images, DICOM metadata and narrative diagnoses. The optimal feature extractors are then integrated into a multimodal representation, which is then clustered to create an automated pipeline for labelling a precursor dataset of 1,337,926 medical images into 50 clusters of visually similar images. The quality of the clusters is assessed by examining their homogeneity and mutual information, taking into account the anatomical region and modality representation. CONCLUSIONS: The results indicate that fusing the embeddings of all three data sources together provides the best results for the task of unsupervised clustering of large-scale medical data and leads to the most concise clusters. Hence, this work marks the initial step towards building a much larger and more fine-grained annotated dataset of medical radiology images.

Long-term motor activity, cardiopulmonary performance and quality of life in abdominal wall defect patients.

AIM: To assess whether patients born with an abdominal wall defect (AWD) have impaired cardiorespiratory performance capacity, motor skills, core stability or quality of life in a long-term follow up. METHODS: Patients diagnosed with AWD between 2002 and 2013 were invited to participate in the study, which included clinical examination, spirometry, cardiopulmonary exercise performance testing, assessment of motor activity, ultrasound, electromyography of the abdominal wall and assessment of the Gastrointestinal Quality of Life Index (GIQLI). The results were compared to a healthy control group matched for age, sex, BMI, and physical activity levels. RESULTS: In total, 18 AWD patients (mean age 12.6 ± 3.5 years) were included and there were no significant differences in anthopometric data compared to the control group (n = 18). AWD patients had a significantly lower GIQLI score (AWD mean 137.2 ± 6.8 vs. control mean 141.4 ± 4.9; p = 0.038) and were affected by decreased motor abilities with significantly higher Dordel-Koch-Test values (AWD median 3.54/IQR 1 vs. control median 2.8/IQR 1; p = 0.005). CONCLUSION: Follow-up examinations of AWD patients revealed decreased motor abilities and GIQLI scores while cardiopulmonary function was not different compared to healthy controls. The clinical impact of these findings remains to be elucidated. IMPACT: Clinical examination, assessment of the gastrointestinal quality of life, sport medical testing, electromyography and abdominal wall ultrasound were performed in patients with congenital abdominal wall defect and compared to an age and sex matched healthy control group. Results of spirometry and spiroergometry, ultrasound or electromyography did not significantly differ between the groups. Significantly decreased locomotor function and gastrointestinal quality of life were found in patients with abdominal wall defect. However, the clinical impact of these findings remains to be elucidated.

The unintended consequences of artificial intelligence in paediatric radiology.

Over the past decade, there has been a dramatic rise in the interest relating to the application of artificial intelligence (AI) in radiology. Originally only 'narrow' AI tasks were possible; however, with increasing availability of data, teamed with ease of access to powerful computer processing capabilities, we are becoming more able to generate complex and nuanced prediction models and elaborate solutions for healthcare. Nevertheless, these AI models are not without their failings, and sometimes the intended use for these solutions may not lead to predictable impacts for patients, society or those working within the healthcare profession. In this article, we provide an overview of the latest opinions regarding AI ethics, bias, limitations, challenges and considerations that we should all contemplate in this exciting and expanding field, with a special attention to how this applies to the unique aspects of a paediatric population. By embracing AI technology and fostering a multidisciplinary approach, it is hoped that we can harness the power AI brings whilst minimising harm and ensuring a beneficial impact on radiology practice.

The "cardiac neglect": a gentle reminder to radiologists interpreting contrast-enhanced abdominal MDCT.

Myocardial infarction (MI) may be visible on contrast-enhanced multidetector computed tomography (MDCT) scans of the abdomen. In the previous literature, potentially missed MI in abdominal MDCTs was not perceived as an issue in radiology. This retrospective single-center study assessed the frequency of detectable myocardial hypoperfusion in contrast-enhanced abdominal MDCTs. We identified 107 patients between 2006 and 2022 who had abdominal MDCTs on the same day or the day before a catheter-proven or clinically evident diagnosis of MI. After reviewing the digital patient records and applying the exclusion criteria, we included 38 patients, with 19 showing areas of myocardial hypoperfusion. All MDCT studies were non ECG-gated. The delay between the MDCT examination and MI diagnosis was shorter in studies with myocardial hypoperfusion (7.4±6.5 hours and 13.8±12.5 hours) but not statistically significant p=0.054. Only 2 of 19 (11%) of these pathologies had been noted in the written radiology reports. The most common cardinal symptom was epigastric pain (50%), followed by polytrauma (21%). STEMI was significantly more common in cases of myocardial hypoperfusion p=0.009. Overall, 16 of 38 (42%) patients died because of acute MI. Based on extrapolations using local MDCT rates, we estimate several thousand radiologically missed MI cases worldwide per year.

Radiological Influencing Factors in the Diagnosis of Painful Habitual Instability of the Thumb Basal Joint as a Precursor of Carpometacarpal Arthritis of the Thumb-A Retrospective Study.

BACKGROUND: Painful habitual instability of the thumb basal joint (PHIT) is a rarely diagnosed condition that can severely impair hand function. Furthermore, it can increase the risk of developing carpometacarpal arthritis of the thumb (CMAOT). Clinical examination and radiographic imaging provide the foundation for a correct diagnosis, but early detection is still challenging. We investigated two objective, radiographically obtainable parameters as potential risk factors for PHIT. METHODS: Clinical data and radiographic images of 33 patients suffering from PHIT were collected and compared to those of 35 people serving as the control group. The two main objectives, the slope angle and the bony offset of the thumb joint, were gathered from the X-rays and statistically analyzed. RESULTS: The analysis showed no differences between the study and the control group concerning the slope angle. Gender and the bony offset, on the other hand, had a significant influence. Female sex and higher offset values were associated with an increased risk of PHIT. CONCLUSIONS: The results of this study prove a connection between a high bony offset and PHIT. We believe this information can be valuable in early detection and will allow more efficient treatment of this condition in the future.

Minimally Invasive Treatment of Trapeziometacarpal Osteoarthritis: Results of a Blinded Randomized Controlled Trial.

BACKGROUND: Surgical therapy for widespread first carpometacarpal (CMC1) arthritis permanently alters the physiologic anatomy of the hand. The injection of autologous substances into the thumb saddle joint could achieve temporary pain relief and delay surgical indications. This trial aimed to compare the pain-reducing effects of autologous fat and/or platelet-rich plasma (PRP) with saline 0.9% in the infiltration therapy of carpometacarpal arthritis of the thumb (CMC1 arthritis). METHODS: A blinded, randomized controlled trial was conducted. Ninety-five patients with CMC1 arthritis were included in the study. The mean follow-up period was 2 years. Participants were injected with 1.5 mL of autologous fat, PRP, autologous fat and PRP, or 0.9% saline solution into the CMC1 joint, depending on the group allocation. The primary outcome of this study was the evaluation of pain reduction in each treatment group. RESULTS: The combination of fat and PRP was the only treatment that resulted in a significantly greater reduction in pain compared with 0.9% saline ( P = 0.003). Similarly, fat and PRP in combination was the only therapy group to achieve clinically relevant Quick Disabilities of Arm, Shoulder, and Hand score reduction, and the only group that showed a significantly better Medical Outcomes Study 36-Item Short-Form Health Survey score than 0.9% saline ( P = 0.016). No major complications were noted. CONCLUSIONS: In addition to pain reduction, the combination of autologous fat and PRP yields a relevant improvement in hand function and a corresponding improvement in quality of life. This treatment seems to be a viable and safe alternative to short-acting glucocorticoids. CLINICAL QUESTION/LEVEL OF EVIDENCE: Therapeutic, II.

The epifascial cap: A typical imaging sign for subcutaneous granuloma annulare in children.

Objectives: Subcutaneous granuloma annulare (SGA) is a rare, self-limiting granulomatous disease in children, commonly diagnosed by histopathology following biopsy or surgical excision. This study aimed to identify imaging clues for SGA that could expedite accurate diagnosis and avoid the need for biopsy in children. Methods: We retrospectively analyzed complete hospital records of all children diagnosed with SGA at our institution from January 2001 to December 2020. Detailed disease history, imaging findings, management, and outcome were evaluated. Results: We identified 28 patients (20 girls) at a median age of 3.75 (range 1-12.5 years). Ten patients presented with multiple lesions. Most lesions were located on the lower extremities (n = 26/41). Ultrasound examinations were performed on all patients, and 12 (43%) patients also received an MRI. Surgical intervention was conducted in 18 (64%) patients either by incisional biopsy (n = 6) or total excision of the lump (n = 12). In all patients who did not undergo surgery, SGA resolved spontaneously. A careful review of the MRIs led to the discovery of a characteristic imaging shape of SGA lesions: the epifascial cap with a typical broad circular base laying on the fascia, extending towards the subdermal/dermal tissue. This distinctive shape was evident in every patient in our cohort. Conclusions: The "Epifascial Cap Sign" is a specific imaging sign for SGA, which to the best of our knowledge, helps distinguish this disease from other subcutaneous lesions. Recognition of this novel diagnostic sign combined with the historical and physical findings should enable clinicians to establish SGA diagnosis easily and diminish the need for further invasive diagnostic procedures.

Distribution and pattern of hand fractures in children and adolescents.

Hand fractures represent commonly encountered injuries in pediatric patients. However, due to modern means of mobility and product safety, the occurrence and distribution of these fractures have changed during the last decades. Therefore, it was the aim of this study to present an update of the epidemiology, pattern, and treatment of hand fractures in a large pediatric cohort. All patients aged between 0 and 17 years treated in our Department in 2019 with fractures of the phalanges, metacarpus, or carpus were included. The medical records were reviewed for age, gender, injury mechanism, fracture localization, season, and treatment. Patients were divided into three different age groups (0-5, 6-12, and 13-17 years). A total of 731 patients with 761 hand fractures were treated during the 1-year study period. The mean age was 11.1 ± 3.5 years, and the majority was male (65%). Male patients were significantly older compared to female patients (p = 0.008). Also, 78.7% of the fractures affected the phalanges, 17.6% the metacarpals, and 3.7% the carpal bones. The proximal phalanges were the most commonly fractured bones (41.5%). Patients with fractures of the carpus were significantly older compared to children sustaining fractures of the metacarpus or phalangeal bones (p < 0.001). Sixteen percent of our patients were treated surgically; these patients were significantly older compared to conservatively treated patients (p = 0.011).  Conclusion: The epidemiology, mechanisms of injury, distribution, and treatment of hand fractures significantly varies among different age groups. This knowledge is of importance for educational purposes of younger colleagues entrusted with care of children and adolescents as well as development of effective prevention strategies. What is Known: • Pediatric hand fractures represent the second most common fractures in children. • The epidemiology of pediatric hand fractures has changed during the last decades and therefore there is a need for an update regarding distribution and epidemiology of pediatric hand fractures. What is New: • In this retrospective cohort study, 761 pediatric hand fractures of 731 patients were analyzed in detail. • The main mechanisms of younger patients were entrapment injuries, older children most commonly sustained their fractures due to ball sport injuries. There was an increasing rate of metacarpal and carpal fractures with increasing age, and these fractures had to be treated operatively more often than phalangeal fractures.

Sonographic Rectal Enema ("hydrocolon") for Diagnosing Large Bowel Pathologies in Infancy - Pictorial Review to Demonstrate Feasibility and Value.

Bowel pathologies encompass a large diversity of diseases with commonly confusing and overlapping clinical presentations. Sonography has a leading role in diagnosing these disorders, especially in small children. However, in some cases, baseline sonography does not deliver a satisfactory result on the suspected pathology. To increase the sensitivity and specificity of the standard bowel ultrasound technique, complimentary ultrasound enema may be performed, which is also referred to as "hydrocolon" in the literature. This paper summarizes the technique of sonographic enema, as well as some bowel pathologies from our case series where sonographic enema proved to be helpful in the diagnostic work-up.

Subcutaneous Granuloma Annulare vs. Subcutaneous Vascular Malformations in Children: A Diagnostic Challenge.

OBJECTIVES: There are various subcutaneous lesions in children and often there is difficulty in obtaining an accurate diagnosis by non-invasive diagnostic procedures. Subcutaneous granuloma annulare (SGA) is a rare granulomatous disease that, even after imaging, is often mistaken for a low-flow subcutaneous vascular malformation (SVM). This study aimed to accurately identify clinical and imaging clues to distinguish SGA from low-flow SVM. METHODS: We retrospectively analyzed complete hospital records of all children with a confirmed diagnosis of SGA and low-flow SVM who underwent MR imaging at our institution from January 2001 to December 2020. Their disease history, clinical and imaging findings, management, and outcome were evaluated. RESULTS: Among 57 patients with granuloma annulare, we identified 12 patients (nine girls) with a confirmed SGA diagnosis who underwent a preoperative MRI. Their median age was 3.25 years (range 2-5 years). Of 455 patients diagnosed with vascular malformations, 90 had malformations limited to the subcutaneous area. Among them only 47 patients with low-flow SVM were included in the study and further analyzed. Our SGA cohort had a female predilection (75%) and a short history of lump appearance of 1.5 months. SGA lesions were immobile and firm. Before MRI, patients underwent initial evaluation by ultrasound (100%) and X-ray (50%). Surgical tissue sampling was performed in all SGA patients to establish a diagnosis. All 47 patients with low-flow SVM were diagnosed correctly by MRI. A total of 45 patients (96%) underwent surgical resection of the SVM. A careful retrospective review of imaging findings of patients with SGA and SVM showed that SGA present as homogenous lesions in the shape of an epifascial cap with a typical broad fascial base extending towards the subdermal tissue in the middle of the lesion. In contrast, SVMs always present with variable-sized multicystic or tubular areas. CONCLUSIONS: Our study shows clear clinical and imaging differences between low-flow SVMs and SGA. SGA presents characteristically in the shape of a homogenous "epifascial cap," which distinguishes these lesions from multicystic heterogenous SVMs.

Information extraction from German radiological reports for general clinical text and language understanding.

Recent advances in deep learning and natural language processing (NLP) have opened many new opportunities for automatic text understanding and text processing in the medical field. This is of great benefit as many clinical downstream tasks rely on information from unstructured clinical documents. However, for low-resource languages like German, the use of modern text processing applications that require a large amount of training data proves to be difficult, as only few data sets are available mainly due to legal restrictions. In this study, we present an information extraction framework that was initially pre-trained on real-world computed tomographic (CT) reports of head examinations, followed by domain adaptive fine-tuning on reports from different imaging examinations. We show that in the pre-training phase, the semantic and contextual meaning of one clinical reporting domain can be captured and effectively transferred to foreign clinical imaging examinations. Moreover, we introduce an active learning approach with an intrinsic strategic sampling method to generate highly informative training data with low human annotation cost. We see that the model performance can be significantly improved by an appropriate selection of the data to be annotated, without the need to train the model on a specific downstream task. With a general annotation scheme that can be used not only in the radiology field but also in a broader clinical setting, we contribute to a more consistent labeling and annotation process that also facilitates the verification and evaluation of language models in the German clinical setting.

Paediatric CT made easy.

Paediatric computed tomography (CT) imaging has always been associated with challenges. Although the technical background of CT imaging is complex, it is worth considering the baseline aspects of radiation exposure to prevent unwanted excess radiation in paediatric patients. In this review, we discuss the most relevant factors influencing radiation exposure, and provide a simplified and practical approach to optimise paediatric CT.

Development and optimization of AI algorithms for wrist fracture detection in children using a freely available dataset.

Introduction: In the field of pediatric trauma computer-aided detection (CADe) and computer-aided diagnosis (CADx) systems have emerged offering a promising avenue for improved patient care. Especially children with wrist fractures may benefit from machine learning (ML) solutions, since some of these lesions may be overlooked on conventional X-ray due to minimal compression without dislocation or mistaken for cartilaginous growth plates. In this article, we describe the development and optimization of AI algorithms for wrist fracture detection in children. Methods: A team of IT-specialists, pediatric radiologists and pediatric surgeons used the freely available GRAZPEDWRI-DX dataset containing annotated pediatric trauma wrist radiographs of 6,091 patients, a total number of 10,643 studies (20,327 images). First, a basic object detection model, a You Only Look Once object detector of the seventh generation (YOLOv7) was trained and tested on these data. Then, team decisions were taken to adjust data preparation, image sizes used for training and testing, and configuration of the detection model. Furthermore, we investigated each of these models using an Explainable Artificial Intelligence (XAI) method called Gradient Class Activation Mapping (Grad-CAM). This method visualizes where a model directs its attention to before classifying and regressing a certain class through saliency maps. Results: Mean average precision (mAP) improved when applying optimizations pre-processing the dataset images (maximum increases of +25.51% mAP@0.5 and +39.78% mAP@[0.5:0.95]), as well as the object detection model itself (maximum increases of +13.36% mAP@0.5 and +27.01% mAP@[0.5:0.95]). Generally, when analyzing the resulting models using XAI methods, higher scoring model variations in terms of mAP paid more attention to broader regions of the image, prioritizing detection accuracy over precision compared to the less accurate models. Discussion: This paper supports the implementation of ML solutions for pediatric trauma care. Optimization of a large X-ray dataset and the YOLOv7 model improve the model's ability to detect objects and provide valid diagnostic support to health care specialists. Such optimization protocols must be understood and advocated, before comparing ML performances against health care specialists.

Bismuth Shielding in Head Computed Tomography-Still Necessary?

Introduction: Cranial CT scans are associated with radiation exposure to the eye lens, which is a particularly radiosensitive organ. Children are more vulnerable to radiation than adults. Therefore, it is essential to use the available dose reduction techniques to minimize radiation exposure. According to the European Consensus on patient contact shielding by the IRCP from 2021, shielding is not recommended in most body areas anymore. This study aims to evaluate whether bismuth shielding as well as its combination with other dose-saving technologies could still be useful. Methods: Cranial CT scans of a pediatric anthropomorphic phantom were performed on two up-to-date MDCT scanners. Eye lens dose measurements were performed using thermoluminescent dosimeters. Furthermore, the impact of BS and of the additional placement of standoff foam between the patient and BS on image quality was also assessed. Results: Bismuth shielding showed a significant lens dose reduction in both CT scanners (GE: 41.50 ± 4.04%, p < 0.001; Siemens: 29.75 ± 6.55%, p = 0.00). When combined with AEC, the dose was lowered even more (GE: 60.75 ± 3.30%, p < 0.001; Siemens: 41.25 ± 8.02%, p = 0.00). The highest eye dose reduction was achieved using BS + AEC + OBTCM (GE: 71.25 ± 2.98%, p < 0.001; Siemens: 58.75 ± 5.85%, p < 0.001). BS caused increased image noise in the orbital region, which could be mitigated by foam placement. Eye shielding had no effect on the image noise in the cranium. Conclusions: The use of BS in cranial CT can lead to a significant dose reduction, which can be further enhanced by its combination with other modern dose reduction methods. BS causes increase in image noise in the orbital region but not in the cranium. The additional use of standoff foam reduces image noise in the orbital region.

Learning rate of students detecting and annotating pediatric wrist fractures in supervised artificial intelligence dataset preparations.

The use of artificial intelligence (AI) in image analysis is an intensively debated topic in the radiology community these days. AI computer vision algorithms typically rely on large-scale image databases, annotated by specialists. Developing and maintaining them is time-consuming, thus, the involvement of non-experts into the workflow of annotation should be considered. We assessed the learning rate of inexperienced evaluators regarding correct labeling of pediatric wrist fractures on digital radiographs. Students with and without a medical background labeled wrist fractures with bounding boxes in 7,000 radiographs over ten days. Pediatric radiologists regularly discussed their mistakes. We found F1 scores-as a measure for detection rate-to increase substantially under specialist feedback (mean 0.61±0.19 at day 1 to 0.97±0.02 at day 10, p<0.001), but not the Intersection over Union as a parameter for labeling precision (mean 0.27±0.29 at day 1 to 0.53±0.25 at day 10, p<0.001). The times needed to correct the students decreased significantly (mean 22.7±6.3 seconds per image at day 1 to 8.9±1.2 seconds at day 10, p<0.001) and were substantially lower as annotated by the radiologists alone. In conclusion our data showed, that the involvement of undergraduated students into annotation of pediatric wrist radiographs enables a substantial time saving for specialists, therefore, it should be considered.

Does Calcium Phosphate Cement Kyphoplasty Cause Intervertebral Disk Degeneration in Adolescents?

OBJECTIVE: Balloon kyphoplasty with polymethylmethacrylate (PMMA) represents the standard procedure for the treatment of thoracic and lumbar type A compression fractures. However, an increased degeneration in adjacent intervertebral disks following PMMA kyphoplasty has been demonstrated in elderly patients. Calcium phosphate cement (CPC) appears to be superior to PMMA for the intravertebral stabilization in younger patients. It remains unkown whether CPC kyphoplasty causes degeneration of adjacent disks in adolescents. DESIGN: Seven adolescents with thoracolumbar spine fractures underwent kyphoplasty at a mean age of 14.5 years (range 10-18). At a mean follow-up of 3.7 years (range 1 to 4.8) postoperatively, 3.0 Tesla magnetic resonance imaging (MRI) of the spine was performed to assess intervertebral disk degeneration by quantitative T2 relaxation maps and subjective ratings using modified Pfirrmann scores. A total of 56 intervertebral disks was analyzed. Initial computed tomography (CT) examinations served as basis to assess the severity of adjacent endplate injuries in terms of articular step-offs. RESULTS: Initial imaging detected 18 thoracolumbar vertebral body fractures of which 9 were treated with CPC kyphoplasty. Quantitative follow-up MRI revealed signs of degeneration in 10 (17.9%) of the examined 56 intervertebral disks, 7 of them adjacent to a previously fractured vertebral body. Signs of disk degeneration were significantly higher in caudal endplates with articular step-offs larger than 5 mm compared to fractured vertebral bodies without endplate step-offs. CONCLUSIONS: Quantitative MRI follow-ups did not suggest CPC-related intervertebral disk degradations following thoracolumbar kyphoplasty in adolescents, but indicated disk alterations correlating to adjacent endplate fracture severity.

Reliability of radiological measurements and treatment recommendations of displaced paediatric humeral medial epicondyle fractures.

There are conflicting recommendations regarding the amount of displacement necessitating stabilization of paediatric humeral medial epicondyle fractures. Our aim was to assess the reliability of the measurements of the displacement and the treatment recommendations of these fractures. The maximum displacement of 57 children with displaced humeral medial epicondyle fractures was analyzed on radiographs by six raters (4 paediatric surgeons, 2 paediatric radiologists) at two time points. In addition, the four surgeons recorded their treatment recommendation. Intraobserver and interobserver reliability were calculated with intraclass correlation coefficients (ICC) and Kappa values. The ICC for the intraobserver reliability ranged between 0.67 and 0.93. The raters disagreed with their own measurements between 8.8% and 28.1%. The ICC for the interobserver reliability of all six raters was 0.90 for measurement 1 and 0.93 for measurement 2. All six raters disagreed (difference > 2mm) in 93% of the cases in measurement 1 and in 91.2% in measurement 2. Treatment recommendations of the four paediatric surgeons between the two time points differed in 5.3% to 28.1% of the cases. Furthermore, the treatment recommendations were concordant in 24 cases (42%) at time point 1 and 32 cases (56.1%) at time point 2. In displaced paediatric medial epicondyle fractures, disagreement regarding measurement of displacement and recommendation for treatment is high. Validated and standardized measurement tools and a clear threshold for operative fixation of displaced medial epicondyle fractures are needed.

Indications and Limitations of Sirolimus in the Treatment of Vascular Anomalies-Insights From a Retrospective Case Series.

Background: Despite recent developments, the role of sirolimus in the heterogeneous spectrum of vascular anomalies is yet to be defined, in terms of indication, dosage, and therapy duration, recognizing both its potential and limitations. Methods: We retrospectively analyzed 16 children with vascular anomalies treated with sirolimus in two pediatric centers between 2014 and 2020 [male: n = 7, the median age at diagnosis: 4.6 months (range, 0-281.4)]. In addition, repetitive volumetric analyses of the vascular anomalies were performed when possible (11 cases). Results: Ten patients were diagnosed with vascular malformations and 6 with vascular tumors. The mean therapy duration was 27.2 months (range, 3.5-65). The mean sirolimus level was 8.52 ng/ml (range, 5.38-12.88). All patients except one with central conducting lymphatic anomaly responded to sirolimus, with the most noticeable volume reduction in the first 4-6 months. Additional administration of vincristine was needed in five patients with kaposiform hemangioendothelioma and yielded a response, even in cases, refractory to sirolimus monotherapy. As a single agent, sirolimus led to impressive improvement in a patient with another vascular tumor-advanced epithelioid hemangioendothelioma. Complicated vascular malformations required long-term sirolimus therapy. Side effects of sirolimus included mucositis and laboratory abnormalities. No major infectious episodes were recorded. An infant with COVID-19, diagnosed while on sirolimus therapy, presented with a mild course. Conclusion: In the current series, we reported limitations of sirolimus as monotherapy, addressing the need to redefine its indications, and explore combination regimens and multimodal treatment strategies. Tools for objective evaluation of response trends over time could serve as a basis for the establishment of future therapeutic algorithms.