Deep Learning for Classification of Bone Lesions on Routine MRI.

BACKGROUND: Radiologists have difficulty distinguishing benign from malignant bone lesions because these lesions may have similar imaging appearances. The purpose of this study was to develop a deep learning algorithm that can differentiate benign and malignant bone lesions using routine magnetic resonance imaging (MRI) and patient demographics. METHODS: 1,060 histologically confirmed bone lesions with T1- and T2-weighted pre-operative MRI were retrospectively identified and included, with lesions from 4 institutions used for model development and internal validation, and data from a fifth institution used for external validation. Image-based models were generated using the EfficientNet-B0 architecture and a logistic regression model was trained using patient age, sex, and lesion location. A voting ensemble was created as the final model. The performance of the model was compared to classification performance by radiology experts. FINDINGS: The cohort had a mean age of 30±23 years and was 58.3% male, with 582 benign lesions and 478 malignant. Compared to a contrived expert committee result, the ensemble deep learning model achieved (ensemble vs. experts): similar accuracy (0·76 vs. 0·73, p=0·7), sensitivity (0·79 vs. 0·81, p=1·0) and specificity (0·75 vs. 0·66, p=0·48), with a ROC AUC of 0·82. On external testing, the model achieved ROC AUC of 0·79. INTERPRETATION: Deep learning can be used to distinguish benign and malignant bone lesions on par with experts. These findings could aid in the development of computer-aided diagnostic tools to reduce unnecessary referrals to specialized centers from community clinics and limit unnecessary biopsies. FUNDING: This work was funded by a Radiological Society of North America Research Medical Student Grant (#RMS2013) and supported by the Amazon Web Services Diagnostic Development Initiative.

Deep learning-based classification of primary bone tumors on radiographs: A preliminary study.

BACKGROUND: To develop a deep learning model to classify primary bone tumors from preoperative radiographs and compare performance with radiologists. METHODS: A total of 1356 patients (2899 images) with histologically confirmed primary bone tumors and pre-operative radiographs were identified from five institutions' pathology databases. Manual cropping was performed by radiologists to label the lesions. Binary discriminatory capacity (benign versus not-benign and malignant versus not-malignant) and three-way classification (benign versus intermediate versus malignant) performance of our model were evaluated. The generalizability of our model was investigated on data from external test set. Final model performance was compared with interpretation from five radiologists of varying level of experience using the Permutations tests. FINDINGS: For benign vs. not benign, model achieved area under curve (AUC) of 0•894 and 0•877 on cross-validation and external testing, respectively. For malignant vs. not malignant, model achieved AUC of 0•907 and 0•916 on cross-validation and external testing, respectively. For three-way classification, model achieved 72•1% accuracy vs. 74•6% and 72•1% for the two subspecialists on cross-validation (p = 0•03 and p = 0•52, respectively). On external testing, model achieved 73•4% accuracy vs. 69•3%, 73•4%, 73•1%, 67•9%, and 63•4% for the two subspecialists and three junior radiologists (p = 0•14, p = 0•89, p = 0•93, p = 0•02, p < 0•01 for radiologists 1-5, respectively). INTERPRETATION: Deep learning can classify primary bone tumors using conventional radiographs in a multi-institutional dataset with similar accuracy compared to subspecialists, and better performance than junior radiologists. FUNDING: The project described was supported by RSNA Research & Education Foundation, through grant number RSCH2004 to Harrison X. Bai.

Age-related fatty infiltration of lumbar paraspinal muscles: a normative reference database study in 516 Chinese females.

BACKGROUND: Fatty infiltration, as a result of aging, is an essential biomarker of muscle degeneration. This research aimed to investigate the age-dependent change of fatty degeneration in the paraspinal muscles of healthy Chinese women. This study also explores the effect of body size on fatty infiltration of paraspinal muscles. METHODS: Cross-sectional area of paraspinal muscles (CSAmuscle) and intermuscular adipose tissue (CSAIMAT) were measured at the L3 mid-vertebral level of 516 healthy females, who underwent abdomen quantitative computed tomography (QCT) scans. Subsequently, IMAT% [CSAIMAT / (CSAIMAT + CSAmuscle)] were calculated. The relationship between basic information and measurements was evaluated using Spearman correlations. Comparisons of QCT results among different BMI subgroups in different age groups were analyzed with the Kruskal-Wallis H test and LSD, post-hoc correction. Age-related changes were calculated after the adjustment of height and weight. RESULTS: The mean CSAIMAT of 20-29 years group (n=69) and 70-79 years group (n=25) were 3.00 cm2 and 11.06 cm2, respectively. While the mean CSAmuscle of 20-29 years group was 38.46 cm2 and 70-79 years group was 30.86 cm2. The mean IMAT% difference between 20-29 years group and 70-79 years group was -18.55%. Strong, positive non-linear associations were observed between ageing and CSAIMAT, along with IMAT% (r=0.656, P<0.01; r=0.714, P<0.01). However, CSAmuscle was shown to decrease with age in a weak, negative linear fashion (r=-0.265, P<0.01). Positive relationships between BMI and CSAIMAT, CSAmuscle, along with IMAT%, were found. Significant differences were observed between obesity and normal BMI subgroup for all variables in three age groups. CSAIMAT showed a larger age-related difference compared to CSAmuscle. CONCLUSIONS: Fatty infiltration in paraspinal muscles increased with age and BMI, while muscle loss may be associated with aging. The present study provided standardized reference data for the fatty degeneration of paraspinal muscles across the adult lifespan of Chinese females, which will play a critical role in future studies.

Radiation/paclitaxel treatment of p53-abnormal non-small cell lung cancer xenograft tumor and associated mechanism.

BACKGROUND: Mutations in key tumor suppressor genes such as tumor protein 53 (TP53) and phosphatase and tensin homolog deleted on chromosome ten (PTEN) are the main genetic alterations in cancers. TP53 mutations have been found in most patients with non-small cell lung cancer (NSCLC), whereas PTEN mutations are rarely found in lung cancer, though most NSCLCs lack PTEN protein synthesis. However, the signaling involved in radio- and chemotherapy of NSCLC with wild-type PTEN and nonfunctional p53 is not clearly understood. METHODS: In this study, we established a xenograft tumor model with H358 NSCLC cells expressing wild-type PTEN, but nonfunctional p53. Protein expression and phosphorylation of PTEN and its downstream signal molecules in NSCLC tissues were detected by Western blot. RESULTS: We demonstrated that radiation and paclitaxel alone inhibited tumor growth, but a combined therapy of radiation and paclitaxel was more effective in inhibiting NSCLC tumor growth. Interestingly, both radiation and paclitaxel significantly increased PTEN protein expression and phosphorylation. Further identification of the affected PTEN downstream molecules showed that Akt phosphorylation at Ser(473) and Thr(308) residues was significantly decreased, whereas Bax and cleaved caspase-3 levels were significantly increased in tumor tissues treated with both radiation and paclitaxel. The combined treatment was more effective than either treatment alone in regulating the studied molecules. We also found that paclitaxel, but not radiation, inhibited phosphoinositide 3-kinase (PI3K) activity. CONCLUSIONS: Our study suggested that a PTEN-PI3K-Akt-Bax signaling cascade is involved in the therapeutic effect of combined radiation/paclitaxel treatment in NSCLC without p53 expression. Our study also suggested that PTEN is an ideal target in tumors with wild-type PTEN and a lack of functional p53.

The mechanism involved in the loss of PTEN expression in NSCLC tumor cells.

Loss of PTEN expression is observed in most non-small cell lung cancers (NSCLC). However, the mechanism by which PTEN expression is regulated in NSCLC has not been fully elucidated. In this study, we investigated the role of DNA methyltransferases (Dnmts), microRNA-29b (miR-29b), and anti-miR-29b inhibitor in PTEN promoter methylation and PTEN gene expression in H358 NSCLC cells in vitro and in vivo. PTEN mRNA was measured by RT-PCR. PTEN and Dnmts protein levels were measured by Western blot. miR-29b expression was detected by Northern blot. A xenograft H358 tumor mouse model was established by subcutaneously inoculating H358 cells into the right hind limbs of nude mice. We found that radiation induced cell apoptosis and hypomethylation in PTEN promoter, PTEN and miR-29b expression, and downregulation of Dnmt1, 3a and 3b expression in H358 tumor cells. The effect of radiation on gene expression and apoptosis was blocked by anti-miR-29b inhibitor. In the xenograft H358 tumor model, anti-miR-29b inhibitor reversed radiation-induced tumor growth delay, PTEN reexpression and downregulation of Dnmts expression. Our study suggested that miR-29b is an upstream molecule of PTEN. miR-29b regulates PTEN gene expression through downregulating Dnmts expression and subsequently induces hypomethylation in PTEN promoter. Targeting therapy could be established in NSCLC by upregulating miR-29b expression.

[Clinical features and imaging findings of intravenous leiomyomatosis].

OBJECTIVE: To explore the clinical features and imaging findings of intravenous leiomyomatosis (IVL). METHOD: The clinical features and imaging findings of 9 patients with pathologically confirmed IVL were retrospectively analyzed. RESULTS: Out of these 9 patients, there were 5 first-episode patients and 4 recurrent patients. Five patients had a history of uterine leiomyoma.Inferior vena cava (IVC) were involved in all 9 patients. The tumor extended through IVC into right heart chamber in 7 patients, among whom tumor were initially arisen from pelvic vessels in 6 patients. The first-episode symptoms included chest tightness and shortness of breath (n=4), edema of low extremity (n=2), abdominal distention (n=2), and menorrhagia (n=1). Tumors in pelvis/venous system and right heart cavity appeared as hypoechoic mass under ultrasound examinations, and hypodense mass with mottled enhancement were observed on contrasted CT. Tumors appeared to be isointense to muscles on T1-weighted images and slightly hyperintense on T2-weighted images. CONCLUSIONS: IVL has certain clinical history and lesion locations. Combined imaging examinations are helpful in the early diagnosis, surgery planning, and follow-up of IVL.

[Application of real-time tissue elastography in diagnosis of breast cancer].

OBJECTIVE: To investigate the clinical value of real-time tissue elastography (RTE) in the diagnosis of breast cancer. METHODS: One hundred and twenty patients with breast lumps (135 lesions) were examined with B-mode imaging, color Doppler flowing imaging (CDFI) and RTE. The elastogram was graded using 5-score evaluating method. The postoperative pathological diagnosis was used as gold standard, and the sensitivity, specificity and accuracy of RTE and two-dimensional ultrasonography combined with RTE in diagnosis of breast cancer were calculated. RESULT: When the score >4 was set for cut-off criteria of malignancy, the sensitivity, specificity and accuracy of RTE was 85.45%, 83.75% and 84.4%, respectively. While two-dimensional ultrasonography combined with RTE was used, the sensitivity, specificity and accuracy increased up to 100%, 95% and 97%, respectively. CONCLUSION: RTE combined with two-dimensional ultrasonography can improve the validity in the diagnosis of malignant breast lesions.