Diagnostic Reliability of Plain Radiography in Osteonecrosis of the Femoral Head: General Radiological Features Revised.

BACKGROUND AND OBJECTIVES: Osteonecrosis of the femoral head (ONFH) is an incapacitating disease that frequently results in the collapse of the femoral head and secondary osteoarthritis. The diagnosis and staging of this pathology, which usually rely on imaging studies, are challenging. Currently, conventional radiography is the basis of the initial diagnostic assessment. In recent decades, however, radiographs have been considered insensitive to early changes in ONFH and thus, a suboptimal diagnostic tool. Paradoxically, the imaging features of radiographs are often profuse, substantial, and characteristic. This study aimed to elucidate the real limitations of this radiologic tool by assessing the diagnostic reliability of the key radiologic features and staging. METHODS: This was a retrospective study in which radiographs from 28 idiopathic ONFH confirmed cases who underwent hip arthroplasty were analyzed by eight observers who were asked to identify the presence or absence of ONFH universally reported imaging features in AP hip radiographs. RESULTS: Concordance analysis revealed a poor agreement between observers for most of the assessed imaging features. Only the identification of femoral head flattening and osteoarthritis signs exhibited moderate agreement with statistical significance. In contrast, the detection of radiological osteoporosis and the loss of trabeculation showed the lowest reliability, with negative kappa coefficients. CONCLUSION: There is a lack of agreement between qualified observers, even for the identification of the most characteristic ONFH radiographic feature. The reliability of plain radiography for the detection of basic radiological elements is even weaker in the early stages of the disease.

Primary intraarterial occlusive lipoma.

Lipomas are the most frequent soft tissue tumors. Intravenous lipomas are very uncommon, but even more unusual are intraarterial lipomas. A 68-year-old heavy smoker man, with chronic alcoholism, retinopathy, dyslipidemia, and a history of type 2 diabetes mellitus of more than 10 years of evolution was hospitalized in a state of dependency. He presented ulcers on both heels and right foot sole extending to the fifth metatarsal base and bedsores in the iliac and sacral regions. Ulcer cultures showed growth of Klebsiella pneumoniae OXA34. A computed tomography angiography scan revealed that the right posterior tibial artery showed several segments with signs of obstruction or sub-occlusive stenosis along its entire length, but especially in the distal two-thirds. The patient underwent supracondylar amputation of the right lower limb. Histopathological sections of the amputated leg showed calcific atherosclerosis obliterans of the posterior tibial artery and complete occlusion in the middle portion of this artery. The occlusion was due to a well-differentiated, white adipose tissue with lipid vacuoles of uniform size. To our knowledge, this case is the first known report of a primary intraarterial lipoma in a peripheral artery. The proliferating adipose tissue within the arterial lumen contributed to distal limb ischemic necrosis. Although an intraarterial lipoma is rare, it should be considered in the differential diagnosis of the causes of peripheral arterial occlusion.

A deep learning model for prognosis prediction after intracranial hemorrhage.

BACKGROUND AND PURPOSE: Intracranial hemorrhage (ICH) is a common life-threatening condition that must be rapidly diagnosed and treated. However, there is still a lack of consensus regarding treatment, driven to some extent by prognostic uncertainty. While several prediction models for ICH detection have already been published, here we present a deep learning predictive model for ICH prognosis. METHODS: We included patients with ICH (n = 262), and we trained a custom model for the classification of patients into poor prognosis and good prognosis, using a hybrid input consisting of brain CT images and other clinical variables. We compared it with two other models, one trained with images only (I-model) and the other with tabular data only (D-model). RESULTS: Our hybrid model achieved an area under the receiver operating characteristic curve (AUC) of .924 (95% confidence interval [CI]: .831-.986), and an accuracy of .861 (95% CI: .760-.960). The I- and D-models achieved an AUC of .763 (95% CI: .622-.902) and .746 (95% CI: .598-.876), respectively. CONCLUSIONS: The proposed hybrid model was able to accurately classify patients into good and poor prognosis. To the best of our knowledge, this is the first ICH prognosis prediction deep learning model. We concluded that deep learning can be applied for prognosis prediction in ICH that could have a great impact on clinical decision-making. Further, hybrid inputs could be a promising technique for deep learning in medical imaging.