Deep learning enables the differentiation between early and late stages of hip avascular necrosis.

OBJECTIVES: To develop a deep learning methodology that distinguishes early from late stages of avascular necrosis of the hip (AVN) to determine treatment decisions. METHODS: Three convolutional neural networks (CNNs) VGG-16, Inception ResnetV2, InceptionV3 were trained with transfer learning (ImageNet) and finetuned with a retrospectively collected cohort of (n = 104) MRI examinations of AVN patients, to differentiate between early (ARCO 1-2) and late (ARCO 3-4) stages. A consensus CNN ensemble decision was recorded as the agreement of at least two CNNs. CNN and ensemble performance was benchmarked on an independent cohort of 49 patients from another country and was compared to the performance of two MSK radiologists. CNN performance was expressed with areas under the curve (AUC), the respective 95% confidence intervals (CIs) and precision, and recall and f1-scores. AUCs were compared with DeLong's test. RESULTS: On internal testing, Inception-ResnetV2 achieved the highest individual performance with an AUC of 99.7% (95%CI 99-100%), followed by InceptionV3 and VGG-16 with AUCs of 99.3% (95%CI 98.4-100%) and 97.3% (95%CI 95.5-99.2%) respectively. The CNN ensemble the same AUCs Inception ResnetV2. On external validation, model performance dropped with VGG-16 achieving the highest individual AUC of 78.9% (95%CI 51.6-79.6%) The best external performance was achieved by the model ensemble with an AUC of 85.5% (95%CI 72.2-93.9%). No significant difference was found between the CNN ensemble and expert MSK radiologists (p = 0.22 and 0.092 respectively). CONCLUSION: An externally validated CNN ensemble accurately distinguishes between the early and late stages of AVN and has comparable performance to expert MSK radiologists. CLINICAL RELEVANCE STATEMENT: This paper introduces the use of deep learning for the differentiation between early and late avascular necrosis of the hip, assisting in a complex clinical decision that can determine the choice between conservative and surgical treatment. KEY POINTS: • A convolutional neural network ensemble achieved excellent performance in distinguishing between early and late avascular necrosis. • The performance of the deep learning method was similar to the performance of expert readers.

Post-mortem CT radiomics for the prediction of time since death.

OBJECTIVES: Post-mortem interval (PMI) estimation has long been relying on sequential post-mortem changes on the body as a function of extrinsic, intrinsic, and environmental factors. Such factors are difficult to account for in complicated death scenes; thus, PMI estimation can be compromised. Herein, we aimed to evaluate the use of post-mortem CT (PMCT) radiomics for the differentiation between early and late PMI. METHODS: Consecutive whole-body PMCT examinations performed between 2016 and 2021 were retrospectively included (n = 120), excluding corpses without an accurately reported PMI (n = 23). Radiomics data were extracted from liver and pancreas tissue and randomly split into training and validation sets (70:30%). Following data preprocessing, significant features were selected (Boruta selection) and three XGBoost classifiers were built (liver, pancreas, combined) to differentiate between early (< 12 h) and late (> 12 h) PMI. Classifier performance was assessed with receiver operating characteristics (ROC) curves and areas under the curves (AUC), which were compared by bootstrapping. RESULTS: A total of 97 PMCTs were included, representing individuals (23 females and 74 males) with a mean age of 47.1 ± 23.38 years. The combined model achieved the highest AUC reaching 75% (95%CI 58.4-91.6%) (p = 0.03 compared to liver and p = 0.18 compared to pancreas). The liver-based and pancreas-based XGBoost models achieved AUCs of 53.6% (95%CI 34.8-72.3%) and 64.3% (95%CI 46.7-81.9%) respectively (p > 0.05 for the comparison between liver- and pancreas-based models). CONCLUSION: The use of radiomics analysis on PMCT examinations differentiated early from late PMI, unveiling a novel image-based method with important repercussions in forensic casework. CLINICAL RELEVANCE STATEMENT: This paper introduces the employment of radiomics in forensic diagnosis by presenting an effective automated alternative method of estimating post-mortem interval from targeted tissues, thus paving the way for improvement in speed and quality of forensic investigations. KEY POINTS: • A combined liver-pancreas radiomics model differentiated early from late post-mortem intervals (using a 12-h threshold) with an area under the curve of 75% (95%CI 58.4-91.6%). • XGBoost models based on liver-only or pancreas-only radiomics demonstrated inferior performance to the combined model in predicting the post-mortem interval.

Pair-Matching Digital 3D Models of Temporomandibular Fragments Using Mesh-To-Mesh Value Comparison and Implications for Commingled Human Remain Assemblages.

The mesh-to-mesh value comparison (MVC) method developed by Karell et al. (Int J Legal Med 130(5):1315-1322, 2016) facilitates the digital comparison of three-dimensional mesh geometries obtained from laser-scanned or computed tomography data of osteological materials. This method has been employed with great success to pair-matching geometries of intact skeletal antimeres, that is, left and right sides. However, as is frequently the case for archaeological materials, there are few circumstances which proffer complete skeletal remains and fewer still when considering contexts of commingling. Prior to the present research, there existed a paucity of sorting techniques for the diverse taphonomic conditions of skeletal materials found within commingled assemblages, especially regarding fragmentary remains. The present chapter details a study in which the MVC method was adapted to encompass comparisons of isolated components of bone in lieu of entire bone geometries in order to address this dearth. Using post-mortem computed tomography data from 35 individuals, three-dimensional models of 70 mandibular fossae and 69 mandibular condyles were created and then compared using Viewbox 4, to produce numerical mesh-to-mesh values which indicate the geometrical and spatial relationship between any two given models. An all-to-all comparison was used to determine if the MVC method, using an automated Trimmed Iterative Closest Point (TrICP) algorithm, could be utilized to (1) match corresponding bilateral pairs of condyles and fossae and (2) match same-sided articular correlates. The pair-matching of both the condyles and the fossae generally produced high sensitivity and specificity rates. However, the articulation results were much poorer and are not currently recommended.

MR Imaging of the Hip: An Update on Bone Marrow Edema.

Bone marrow edema (BME) represents a frequent and easily detected MR imaging finding that is recognized as a common source of pain. It is considered a nonspecific finding that constitutes a central component of a wide spectrum of pathologies affecting the hip joint, showing high variability in terms of prognosis and treatment requirements. Solid knowledge of the various BME patterns and associated imaging findings indicative of these conditions can aid in differentiation. Additionally, correlation with clinical and laboratory data are essential for arriving at a final diagnosis in several cases. In this review, we describe the disease characteristics and imaging findings and also highlight the differential diagnostic clues regarding common hip disorders associated with BME including avascular necrosis, transient osteoporosis, stress injuries, and infectious/inflammatory disorders.

Three-Dimensional Geometry of Phalanges as a Proxy for Pair-Matching: Mesh Comparison Using an ICP Algorithm.

Forensic anthropologists are frequently faced with the challenge of individualizing and sorting commingled remains in a variety of scenarios. A number of protocols have been proposed to standardize the methodological approach to individuating commingled remains, some of which are focused on pair-matching. A recent study by Karell et al. (2016) proposed a virtual method for pair-matching humeri using a semi-automatic procedure that gave encouraging results. With regards to the phalanges, there are only a handful of studies focusing on identifying and siding phalanges, as well as exploring their directional and functional asymmetry. Yet, they are still as important as every other bone when sorting commingled human remains in various situations, such as archaeological common burials and mass graves, commingled decomposed remains resulting from atrocities, accidents or natural disasters. This study investigates a new method for pair-matching, a common individualization technique, using digital three-dimensional models of bone: mesh-to-mesh value comparison (MVC) as proposed by Karell et al. (2016). The MVC method digitally compares the entire three-dimensional geometry of two bones using an iterative closest point (ICP) algorithm to produce a single value as a proxy for their similarity. The method is automated with the use of Viewbox software 4.1 beta for a simultaneous comparison of all possible pairs. For this study, 515 phalanges from 24 individuals of mixed ancestry were digitized using CT scans and the 3D modeling program AMIRA 5.3.3. The models were also hollowed (internal information of compact and trabecular bone removed) to test the method with simulated surface scan models. The subsequent data-over 73,000 comparisons-were assessed using sensitivity and specificity rates via ROC analysis to indicate how well the automated version of MVC pair-matched phalanges. The best bone in terms of pair-matching was the proximal phalanx of Digit 3 with 87.5% sensitivity and 92.4% specificity rates at a threshold value of 0.488 for the unhollowed bones. The specificity drops slightly (91.1%) when the hollowed models are compared. To compare the performance of the method in all phalanges, the specificity was set to 95%-allowing for a 5% acceptable error-and the adjusted sensitivity was compared. The highest sensitivity, namely 68.8%, was noted for Digit 2 proximal phalanx for both unhollowed and hollowed models. Thus far, our preliminary results indicate that the MVC method performs well when pair-matching phalanges, though it is less accurate than pair-matching other types of bones. The introduction of 95% specificity threshold allows for rejecting pairs in great confidence, which could, for instance, significantly reduce the number of DNA comparisons required for the remaining possible matches. In addition, the similar results obtained from hollowed and unhollowed models indicate that the internal information included in the unhollowed models adds little to the identification of true pairs. This means that if a CT scan is not available, the method could be applied to surface models produced by light and laser scanners as well. While additional work needs to be done to verify these preliminary results, this research has the potential to expand the repertoire of individualization methods.

Multifocal bone and bone marrow lesions in children - MRI findings.

Polyostotic bone and bone marrow lesions in children may be due to various disorders. Radiographically, lytic lesions may become apparent after loss of more than 50% of the bone mineral content. Scintigraphy requires osteoblastic activity and is not specific. MRI may significantly contribute to the correct diagnosis and management. Accurate interpretation of MRI examinations requires understanding of the normal conversion pattern of bone marrow in childhood and of the appearances of red marrow rests and hyperplasia. Differential diagnosis is wide: Malignancies include metastases, multifocal primary sarcomas and hematological diseases. Benign entities include benign tumors and tumor-like lesions, histiocytosis, infectious and inflammatory diseases, multiple stress fractures/reactions and bone infarcts/ischemia.

Sexual dimorphism of the bony labyrinth: a new age-independent method.

Currently in physical anthropology there is a need for reliable methods of sex estimation for immature individuals and highly fragmented remains. This study develops a sex estimation technique from discriminant function analysis of the bony labyrinth as it matures before puberty and can survive taphonomic conditions that would destroy most other skeletal material. The bony labyrinth contains the organs of hearing and balance. For this reason biologists and paleoanthropologists have undertaken research in this area to understand evolutionary changes in locomotion. Prior studies have found clear differences between species, but within-species variation has not been satisfactorily investigated. 3D segmentations of the left and right labyrinths of 94 individuals from a Cretan collection were generated and measured. Mean measurements of height, width, size, and shape indices were analyzed for sexual dimorphism, bilateral asymmetry, and measurement error. Significant sexual dimorphism was detected for several measurements. For sex estimation, the single best variable was the radius of curvature of the posterior semicircular canal, which achieved 76% accuracy. Two multivariate functions increased accuracy to 84%. Although these equations are less accurate than equations for complete long bones and crania, they appear to be as accurate as or better than other techniques for sexing immature individuals and temporal bones.

Radiomics for the Detection of Active Sacroiliitis Using MR Imaging.

Detecting active inflammatory sacroiliitis at an early stage is vital for prescribing medications that can modulate disease progression and significantly delay or prevent debilitating forms of axial spondyloarthropathy. Conventional radiography and computed tomography offer limited sensitivity in detecting acute inflammatory findings as these methods primarily identify chronic structural lesions. Conversely, Magnetic Resonance Imaging (MRI) is the preferred technique for detecting bone marrow edema, although it is a complex process requiring extensive expertise. Additionally, ascertaining the origin of lesions can be challenging, even for experienced medical professionals. Machine learning (ML) has showcased its proficiency in various fields by uncovering patterns that are not easily perceived from multi-dimensional datasets derived from medical imaging. The aim of this study is to develop a radiomic signature to aid clinicians in diagnosing active sacroiliitis. A total of 354 sacroiliac joints were segmented from axial fluid-sensitive MRI images, and their radiomic features were extracted. After selecting the most informative features, a number of ML algorithms were utilized to identify the optimal method for detecting active sacroiliitis, leading to the selection of an Extreme Gradient Boosting (XGBoost) model that accomplished an Area Under the Receiver-Operating Characteristic curve (AUC-ROC) of 0.71, thus further showcasing the potential of radiomics in the field.

Retroperitoneal Soft Tissue Sarcoma: Emerging Therapeutic Strategies.

Retroperitoneal soft tissue sarcoma (RPS) is a rare and heterogenous disease for which surgery is the cornerstone of treatment. However, the local recurrence rate is much higher than in soft tissue sarcoma of the extremities since wide resection is usually unfeasible in RPS due to its large size, indistinct tumour borders, anatomical constraints and the thinness of the overlying peritoneum. Local recurrence is the leading cause of death for low-grade RPS, whereas high-grade tumours are prone to distant metastases. In recent decades, the role of emerging therapeutic strategies, such as more extended surgery and (neo)adjuvant treatments to improve oncological outcome in primary localised RPS, has been extensively investigated. In this review, the recent data on the evolving multidisciplinary management of primary localised RPS are comprehensively discussed. The heterogeneity of RPS, with their different histological subtypes and biological behaviour, renders a standard therapeutic 'one-size-fits-all' approach inappropriate, and treatment should be modified according to histological type and malignancy grade. There is sufficient evidence that frontline extended surgery with compartmental resection including all ipsilateral retroperitoneal fat and liberal en bloc resection of adjacent organs and structures, even if they are not macroscopically involved, increases local tumour control in low-grade sarcoma and liposarcoma, but not in leiomyosarcoma for which complete macroscopic resection seems sufficient. Additionally, preoperative radiotherapy is not indicated for all RPSs, but seems to be beneficial in well-differentiated liposarcoma and grade I/II dedifferentiated liposarcoma, and probably in solitary fibrous tumour. Whether neoadjuvant chemotherapy is of benefit in high-grade RPS remains unclear from retrospective data and is subject of the ongoing randomised STRASS 2 trial, from which the results are eagerly awaited. Personalised, histology-tailored multimodality treatment is promising and will likely further evolve as our understanding of the molecular and genetic characteristics within RPS improves.

Deep Learning for the Differential Diagnosis between Transient Osteoporosis and Avascular Necrosis of the Hip.

Differential diagnosis between avascular necrosis (AVN) and transient osteoporosis of the hip (TOH) can be complicated even for experienced MSK radiologists. Our study attempted to use MR images in order to develop a deep learning methodology with the use of transfer learning and a convolutional neural network (CNN) ensemble, for the accurate differentiation between the two diseases. An augmented dataset of 210 hips with TOH and 210 hips with AVN was used to finetune three ImageNet-trained CNNs (VGG-16, InceptionResNetV2, and InceptionV3). An ensemble decision was reached in a hard-voting manner by selecting the outcome voted by at least two of the CNNs. Inception-ResNet-V2 achieved the highest AUC (97.62%) similar to the model ensemble, followed by InceptionV3 (AUC of 96.82%) and VGG-16 (AUC 96.03%). Precision for the diagnosis of AVN and recall for the detection of TOH were higher in the model ensemble compared to Inception-ResNet-V2. Ensemble performance was significantly higher than that of an MSK radiologist and a fellow (P < 0.001). Deep learning was highly successful in distinguishing TOH from AVN, with a potential to aid treatment decisions and lead to the avoidance of unnecessary surgery.

Differentiation between subchondral insufficiency fractures and advanced osteoarthritis of the knee using transfer learning and an ensemble of convolutional neural networks.

PURPOSE: Subchondral insufficiency fractures (SIF) and advanced osteoarthritis (OA) of the knee are usually seen in conjunction with bone marrow lesions (BMLs) and their differentiation may pose a significant diagnostic challenge. We aimed to develop a convolutional neural network (CNN) ensemble which could successfully differentiate between these two entities. MATERIALS AND METHODS: A total of 212 knees with SIF and 102 knees with advanced OA with BMLs were retrospectively included. Coronal fat suppressed PD-w images were augmented, resized and normalized, reaching a total of 1174 images. Data was used to fine-tune three ImageNet-pretrained CNNs (VGG-16, InceptionV3 and Inception-ResNet-V2). Agreement of at least two networks was recorded as the decision of the network ensemble. Ensemble performance was compared to that of two MSK radiologists on the validation set. Receiver operating characteristics (ROC) curves and the respective areas under the curve (AUC) were used to evaluate human and machine performance. RESULTS: InceptionV3 achieved the highest AUC (93.68%) and VGG-16 the lowest AUC (82.18%) among individual CNNs. CNN ensemble achieved the highest overall performance with an AUC of 95.97%. The first of the two MSK radiologists achieved a performance similar to the ensemble, reaching an AUC of 91.95%. The second radiologist achieved lower AUC of 82.76% which was lower than both the other specialist and the ensemble (P < 0.001). CONCLUSION: A CNN ensemble was highly accurate in differentiating between SIF and OA, achieving a higher or equal performance to MSK radiologists.

Postnatal changes of the human bony labyrinth morphology.

It has been suggested that the inner ear attains its final morphology already in utero and that environmental factors do not influence its size or shape after birth. Thus, direct comparison between the adult and the subadult bony labyrinths can be made. Herein, this hypothesis was tested in a sample of 170 subadult individuals. Three-dimensional (3D) models of the labyrinths were created for 170 French and Greek juveniles. Ten linear measurements of the 3D models were taken and the labyrinthine shape was evaluated by calculating 10 shape indices. All variables were subsequently analyzed in the context of postnatal development. Significant positive correlation between the chronological age and several variables was noted suggesting that age-related changes of the bony labyrinth occur in the postnatal period. Our findings contradict the simplified assumption that the bony labyrinth retains a stable morphology after birth. Further research should concentrate on evaluating postnatal changes to the temporal bone and their influence on the labyrinthine morphology.

Solitary upper thoracic vertebral metastasis as an initial presentation of pancreatic cancer: A case report.

The spine represents an unusual site of pancreatic metastatic disease, while the exact incidence of this metastatic lesion is unclear. An extremely rare case of bone painful blastic metastatic lesion at the fourth thoracic vertebra, as a first manifestation of pancreatic cancer, is reported in the current study. A 54-year-old man was complaining of upper thoracic spinal pain that lasted the previous 4 months. A CT scan revealed a solitary, sclerotic, bone lesion of the body of the fourth thoracic (T4) vertebra, indicating a benign lesion. A consequent MRI scan of the thoracic spine confirmed the sclerotic lesion along with additional findings, such as bone marrow edema of the vertebra and a paraspinal and epidural soft tissue mass. Biopsy of the T4 vertebra lesion revealed metastatic lesion of pancreatic origin. A CT scan of the upper abdomen confirmed the diagnosis. This is a rare case, since the only initial symptom of pancreatic cancer was pain of the upper spine, while the first imaging was misleading, indicating a lesion that was benign in nature.

Unenhanced PMCT in the diagnosis of fatal traumatic brain injury in a charred body.

Traffic incidents are one of the most frequent causes of death in young adults worldwide. Depending on the location of the incident, the velocity of the vehicle(s), the weather condition, traffic incidents are often complicated making the investigation of the circumstances difficult. Here we report a case of an incinerated body involved in a vehicle incident. Differential diagnosis included natural cause of death during driving, fatal traumatic injuries, death due to fire and positional asphyxia. The body was submitted to PMCT prior to autopsy as part of a research protocol (N. 1388/2016) at the Department of Medical Imaging of the University Hospital of Heraklion in Crete, Greece. Unenhanced PMCT revealed craniofacial fractures, a thin film of subdural haemorrhage and an epidural fluid collection. The findings were interpreted as consistent with an impact to the face, causing craniofacial fractures mainly on the right side, and an acute subdural hematoma. Autopsy findings corroborated the diagnosis. The epidural hematoma was deemed to be post-mortem heat-induced. This case is an excellent example of the diagnostic value of PMCT in the medicolegal investigation of death.

Radiomics and Machine Learning Can Differentiate Transient Osteoporosis from Avascular Necrosis of the Hip.

Differentiation between transient osteoporosis (TOH) and avascular necrosis (AVN) of the hip is a longstanding challenge in musculoskeletal radiology. The purpose of this study was to utilize MRI-based radiomics and machine learning (ML) for accurate differentiation between the two entities. A total of 109 hips with TOH and 104 hips with AVN were retrospectively included. Femoral heads and necks with segmented radiomics features were extracted. Three ML classifiers (XGboost, CatBoost and SVM) using 38 relevant radiomics features were trained on 70% and validated on 30% of the dataset. ML performance was compared to two musculoskeletal radiologists, a general radiologist and two radiology residents. XGboost achieved the best performance with an area under the curve (AUC) of 93.7% (95% CI from 87.7 to 99.8%) among ML models. MSK radiologists achieved an AUC of 90.6% (95% CI from 86.7% to 94.5%) and 88.3% (95% CI from 84% to 92.7%), respectively, similar to residents. The general radiologist achieved an AUC of 84.5% (95% CI from 80% to 89%), significantly lower than of XGboost (p = 0.017). In conclusion, radiomics-based ML achieved a performance similar to MSK radiologists and significantly higher compared to general radiologists in differentiating between TOH and AVN.

Spectrum of skeletal disorders during the peripartum period: MRI patterns.

Pregnancy is a normal condition in which a combination of biomechanical and hormonal changes may result in a wide spectrum of skeletal disorders. Skeletal overloading due to postural and weight changes, combined with increased ligamentous laxity stemming from the action of relaxing and the risk of pregnancy-related osteoporosis during childbearing, childbirth, and postpartum, have been associated with various pathologies. Owing to the heterogeneity of proposed contributing factors, skeletal lesions in the peripartum period may be related to different pathogenesis depending on the metabolic status of the patient and the degree of biomechanical stress; thus, a fatigue or insufficiency discrimination is not always easy to make. When combined with clinical data, magnetic resonance imaging (MRI) is a powerful tool in clarifying the cause of skeletal pain in the majority of patients or alerting clinicians to proceed to further investigations. Early detection and tailored treatment are important in order to avoid disease progression and long-term restriction of daily activities which may have an adverse impact on the relationship between the mother and the newborn. This pictorial essay provides an overview of the MRI characteristics and pattern of involvement of skeletal lesions presenting during the peripartum period, combined with demographic data and dual-energy X-ray absorptiometry (DEXA) measurements.