Artificial intelligence solution to classify pulmonary nodules on CT.

PURPOSE: The purpose of this study was to create an algorithm to detect and classify pulmonary nodules in two categories based on their volume greater than 100 mm3 or not, using machine learning and deep learning techniques. MATERIALS AND METHOD: The dataset used to train the model was provided by the organization team of the SFR (French Radiological Society) Data Challenge 2019. An asynchronous and parallel 3-stages pipeline was developed to process all the data (a data "pre-processing" stage; a "nodule detection" stage; a "classifier" stage). Lung segmentation was achieved using 3D U-NET algorithm; nodule detection was done using 3D Retina-UNET and classifier stage with a support vector machine algorithm on selected features. Performances were assessed using area under receiver operating characteristics curve (AUROC). RESULTS: The pipeline showed good performance for pathological nodule detection and patient diagnosis. With the preparation dataset, an AUROC of 0.9058 (95% confidence interval [CI]: 0.8746-0.9362) was obtained, 87% yielding accuracy (95% CI: 84.83%-91.03%) for the "nodule detection" stage, corresponding to 86% specificity (95% CI: 82%-92%) and 89% sensitivity (95% CI: 84.83%-91.03%). CONCLUSION: A fully functional pipeline using 3D U-NET, 3D Retina-UNET and classifier stage with a support vector machine algorithm was developed, resulting in high capabilities for pulmonary nodule classification.

Artificial intelligence to predict clinical disability in patients with multiple sclerosis using FLAIR MRI.

PURPOSE: The purpose of this study was to create an algorithm that combines multiple machine-learning techniques to predict the expanded disability status scale (EDSS) score of patients with multiple sclerosis at two years solely based on age, sex and fluid attenuated inversion recovery (FLAIR) MRI data. MATERIALS AND METHODS: Our algorithm combined several complementary predictors: a pure deep learning predictor based on a convolutional neural network (CNN) that learns from the images, as well as classical machine-learning predictors based on random forest regressors and manifold learning trained using the location of lesion load with respect to white matter tracts. The aggregation of the predictors was done through a weighted average taking into account prediction errors for different EDSS ranges. The training dataset consisted of 971 multiple sclerosis patients from the "Observatoire français de la sclérose en plaques" (OFSEP) cohort with initial FLAIR MRI and corresponding EDSS score at two years. A test dataset (475 subjects) was provided without an EDSS score. Ten percent of the training dataset was used for validation. RESULTS: Our algorithm predicted EDSS score in patients with multiple sclerosis and achieved a MSE=2.2 with the validation dataset and a MSE=3 (mean EDSS error=1.7) with the test dataset. CONCLUSION: Our method predicts two-year clinical disability in patients with multiple sclerosis with a mean EDSS score error of 1.7, using FLAIR sequence and basic patient demographics. This supports the use of our model to predict EDSS score progression. These promising results should be further validated on an external validation cohort.

Abdominal musculature segmentation and surface prediction from CT using deep learning for sarcopenia assessment.

PURPOSE: The purpose of this study was to build and train a deep convolutional neural networks (CNN) algorithm to segment muscular body mass (MBM) to predict muscular surface from a two-dimensional axial computed tomography (CT) slice through L3 vertebra. MATERIALS AND METHODS: An ensemble of 15 deep learning models with a two-dimensional U-net architecture with a 4-level depth and 18 initial filters were trained to segment MBM. The muscular surface values were computed from the predicted masks and corrected with the algorithm's estimated bias. Resulting mask prediction and surface prediction were assessed using Dice similarity coefficient (DSC) and root mean squared error (RMSE) scores respectively using ground truth masks as standards of reference. RESULTS: A total of 1025 individual CT slices were used for training and validation and 500 additional axial CT slices were used for testing. The obtained mean DSC and RMSE on the test set were 0.97 and 3.7 cm2 respectively. CONCLUSION: Deep learning methods using convolutional neural networks algorithm enable a robust and automated extraction of CT derived MBM for sarcopenia assessment, which could be implemented in a clinical workflow.

Three artificial intelligence data challenges based on CT and MRI.

PURPOSE: The second edition of the artificial intelligence (AI) data challenge was organized by the French Society of Radiology with the aim to: (i), work on relevant public health issues; (ii), build large, multicentre, high quality databases; and (iii), include three-dimensional (3D) information and prognostic questions. MATERIALS AND METHODS: Relevant clinical questions were proposed by French subspecialty colleges of radiology. Their feasibility was assessed by experts in the field of AI. A dedicated platform was set up for inclusion centers to safely upload their anonymized examinations in compliance with general data protection regulation. The quality of the database was checked by experts weekly with annotations performed by radiologists. Multidisciplinary teams competed between September 11th and October 13th 2019. RESULTS: Three questions were selected using different imaging and evaluation modalities, including: pulmonary nodule detection and classification from 3D computed tomography (CT), prediction of expanded disability status scale in multiple sclerosis using 3D magnetic resonance imaging (MRI) and segmentation of muscular surface for sarcopenia estimation from two-dimensional CT. A total of 4347 examinations were gathered of which only 6% were excluded. Three independent databases from 24 individual centers were created. A total of 143 participants were split into 20 multidisciplinary teams. CONCLUSION: Three data challenges with over 1200 general data protection regulation compliant CT or MRI examinations each were organized. Future challenges should be made with more complex situations combining histopathological or genetic information to resemble real life situations faced by radiologists in routine practice.

Long-term recurrences of jaw osteoradionecrosis after apparent healing with the PENTOCLO protocol.

Osteoradionecrosis of the jaws (ORNJ) is a late complication of head and neck irradiation estimated at around 3% of irradiated patients. The PENTO protocol (Pentoxyfilline and Tocopherol), with the eventual adjunction of Clodronate (PENTOCLO), showed interesting results even in advanced ORNJ. The current literature does not describe the long-term outcomes and particularly after the completion of the protocol. The PENTO or PENTOCLO protocol should be prescribed as a life-long treatment or the outcome should be monitored at least as long as the duration of the protocol after its end.

Diagnosis of focal liver lesions from ultrasound using deep learning.

PURPOSE: The purpose of this study was to create an algorithm that simultaneously detects and characterizes (benign vs. malignant) focal liver lesion (FLL) using deep learning. MATERIALS AND METHODS: We trained our algorithm on a dataset proposed during a data challenge organized at the 2018 Journées Francophones de Radiologie. The dataset was composed of 367 two-dimensional ultrasound images from 367 individual livers, captured at various institutions. The algorithm was guided using an attention mechanism with annotations made by a radiologist. The algorithm was then tested on a new data set from 177 patients. RESULTS: The models reached mean ROC-AUC scores of 0.935 for FLL detection and 0.916 for FLL characterization over three shuffled three-fold cross-validations performed with the training data. On the new dataset of 177 patients, our models reached a weighted mean ROC-AUC scores of 0.891 for seven different tasks. CONCLUSION: This study that uses a supervised-attention mechanism focused on FLL detection and characterization from liver ultrasound images. This method could prove to be highly relevant for medical imaging once validated on a larger independent cohort.

Kidney cortex segmentation in 2D CT with U-Nets ensemble aggregation.

PURPOSE: This work presents our contribution to one of the data challenges organized by the French Radiology Society during the Journées Francophones de Radiologie. This challenge consisted in segmenting the kidney cortex from coronal computed tomography (CT) images, cropped around the cortex. MATERIALS AND METHODS: We chose to train an ensemble of fully-convolutional networks and to aggregate their prediction at test time to perform the segmentation. An image database was made available in 3 batches. A first training batch of 250 images with segmentation masks was provided by the challenge organizers one month before the conference. An additional training batch of 247 pairs was shared when the conference began. Participants were ranked using a Dice score. RESULTS: The segmentation results of our algorithm match the renal cortex with a good precision. Our strategy yielded a Dice score of 0.867, ranking us first in the data challenge. CONCLUSION: The proposed solution provides robust and accurate automatic segmentations of the renal cortex in CT images although the precision of the provided reference segmentations seemed to set a low upper bound on the numerical performance. However, this process should be applied in 3D to quantify the renal cortex volume, which would require a marked labelling effort to train the networks.

Five simultaneous artificial intelligence data challenges on ultrasound, CT, and MRI.

PURPOSE: The goal of this data challenge was to create a structured dynamic with the following objectives: (1) teach radiologists the new rules of General Data Protection Regulation (GDPR), while building a large multicentric prospective database of ultrasound, computed tomography (CT) and MRI patient images; (2) build a network including radiologists, researchers, start-ups, large companies, and students from engineering schools, and; (3) provide all French stakeholders working together during 5 data challenges with a secured framework, offering a realistic picture of the benefits and concerns in October 2018. MATERIALS AND METHODS: Relevant clinical questions were chosen by the Société Francaise de Radiologie. The challenge was designed to respect all French ethical and data protection constraints. Multidisciplinary teams with at least one radiologist, one engineering student, and a company and/or research lab were gathered using different networks, and clinical databases were created accordingly. RESULTS: Five challenges were launched: detection of meniscal tears on MRI, segmentation of renal cortex on CT, detection and characterization of liver lesions on ultrasound, detection of breast lesions on MRI, and characterization of thyroid cartilage lesions on CT. A total of 5,170 images within 4 months were provided for the challenge by 46 radiology services. Twenty-six multidisciplinary teams with 181 contestants worked for one month on the challenges. Three challenges, meniscal tears, renal cortex, and liver lesions, resulted in an accuracy>90%. The fourth challenge (breast) reached 82% and the lastone (thyroid) 70%. CONCLUSION: Theses five challenges were able to gather a large community of radiologists, engineers, researchers, and companies in a very short period of time. The accurate results of three of the five modalities suggest that artificial intelligence is a promising tool in these radiology modalities.

French updated recommendations in Stage I to III melanoma treatment and management.

As knowledge continues to develop, regular updates are necessary concerning recommendations for practice. The recommendations for the management of melanoma stages I to III were drawn up in 2005. At the request of the Société Française de Dermatologie, they have now been updated using the methodology for recommendations proposed by the Haute Autorité de Santé in France. In practice, the principal recommendations are as follows: for staging, it is recommended that the 7th edition of AJCC be used. The maximum excision margins have been reduced to 2 cm. Regarding adjuvant therapy, the place of interferon has been reduced and no validated emerging medication has yet been identified. Radiotherapy may be considered for patients in Stage III at high risk of relapse. The sentinel lymph node technique remains an option. Initial examination includes routine ultrasound as of Stage II, with other examinations being optional in stages IIC and III. A shorter strict follow-up period (3 years) is recommended for patients, but with greater emphasis on imaging.

Selection of an early biomarker for vascular normalization using dynamic contrast-enhanced ultrasonography to predict outcomes of metastatic patients treated with bevacizumab.

BACKGROUND: Dynamic contrast-enhanced ultrasonography (DCE-US) has been used for evaluation of tumor response to antiangiogenic treatments. The objective of this study was to assess the link between DCE-US data obtained during the first week of treatment and subsequent tumor progression. PATIENTS AND METHODS: Patients treated with antiangiogenic therapies were included in a multicentric prospective study from 2007 to 2010. DCE-US examinations were available at baseline and at day 7. For each examination, a 3 min perfusion curve was recorded just after injection of a contrast agent. Each perfusion curve was modeled with seven parameters. We analyzed the correlation between criteria measured up to day 7 on freedom from progression (FFP). The impact was assessed globally, according to tumor localization and to type of treatment. RESULTS: The median follow-up was 20 months. The mean transit time (MTT) evaluated at day 7 was the only criterion significantly associated with FFP (P = 0.002). The cut-off point maximizing the difference between FFP curves was 12 s. Patients with at least a 12 s MTT had a better FFP. The results according to tumor type were significantly heterogeneous: the impact of MTT on FFP was more marked for breast cancer (P = 0.004) and for colon cancer (P = 0.025) than for other tumor types. Similarly, the differences in FFP according to MTT at day 7 were marked (P = 0.004) in patients receiving bevacizumab. CONCLUSION: The MTT evaluated with DCE-US at day 7 is significantly correlated to FFP of patients treated with bevacizumab. This criterion might be linked to vascular normalization. AFSSAPS NO: 2007-A00399-44.

[Update to the recommendations for management of melanoma stages I to III]. / Actualisation des recommandations de prise en charge du mélanome stades I à III.

As knowledge continues to develop, regular updates are necessary concerning recommendations for practice. The recommendations for the management of melanoma stages I to III were drawn up in 2005. At the request of the Société Française de Dermatologie, they have now been updated using the methodology for recommendations proposed by the Haute Autorité de Santé. In practice, the principal recommendations are as follows: for staging, it is recommended that the 7th edition of AJCC be used. The maximum excision margins have been reduced to 2cm. Regarding adjuvant therapy, the place of interferon has been reduced and no validated emerging medication has yet been identified. Radiotherapy may be considered for patients in stage III at high risk of relapse. The sentinel lymph node technique remains an option. Initial examination includes routine ultrasound as of stage II, with other examinations being optional in stages IIC and III. A shorter strict follow-up period (3years) is recommended for patients, but with greater emphasis on imaging.

Guidelines and good clinical practice recommendations for contrast enhanced ultrasound (CEUS) in the liver--update 2012: a WFUMB-EFSUMB initiative in cooperation with representatives of AFSUMB, AIUM, ASUM, FLAUS and ICUS.

Initially, a set of guidelines for the use of ultrasound contrast agents was published in 2004 dealing only with liver applications. A second edition of the guidelines in 2008 reflected changes in the available contrast agents and updated the guidelines for the liver, as well as implementing some non-liver applications. Time has moved on, and the need for international guidelines on the use of CEUS in the liver has become apparent. The present document describes the third iteration of recommendations for the hepatic use of contrast enhanced ultrasound (CEUS) using contrast specific imaging techniques. This joint WFUMB-EFSUMB initiative has implicated experts from major leading ultrasound societies worldwide. These liver CEUS guidelines are simultaneously published in the official journals of both organizing federations (i.e., Ultrasound in Medicine and Biology for WFUMB and Ultraschall in der Medizin/European Journal of Ultrasound for EFSUMB). These guidelines and recommendations provide general advice on the use of all currently clinically available ultrasound contrast agents (UCA). They are intended to create standard protocols for the use and administration of UCA in liver applications on an international basis and improve the management of patients worldwide.

An EFSUMB introduction into Dynamic Contrast-Enhanced Ultrasound (DCE-US) for quantification of tumour perfusion.

Dynamic Contrast Enhanced Ultrasound (DCE-US) is an imaging technique that utilizes microbubble contrast agents in diagnostic ultrasound. The EFSUMB guidelines published in 2004, updated in 2008 and 2011 focused on the use of DCE-US, including essential technical requirements, training, investigational procedures and steps, guidance on image interpretation, established and recommended clinical indications and safety considerations. However the quantification of images acquired with ultrasound contrast agents (UCAs) is not discussed in the guidelines. The purpose of this EFSUMB document is to provide some recommendations and descriptions of the quantification of ultrasound images, technical requirements for analysis of time-intensity curves (TICs), methodology for data analysis, and interpretation of the results.

Sequential research-related biopsies in phase I trials: acceptance, feasibility and safety.

BACKGROUND: Sequential tumour biopsies are of potential interest for the rational development of molecular targeted therapies. PATIENTS AND METHODS: From June 2004 to July 2009, 186 patients participated in 14 phase I clinical trials in which sequential tumour biopsies (13 trials) and/or sequential normal skin biopsies (6 trials) were optional. All patients had to sign an independent informed consent for the biopsies. RESULTS: Tumour biopsies were proposed to 155 patients and 130 (84%) signed the consent while normal skin biopsies were proposed to 70 patients and 57 (81%) signed the consent. Tumour biopsies could not be carried out in 41 (31%) of the 130 consenting patients. Tumour biopsies were collected at baseline in 33 patients, at baseline and under treatment in 56 patients. Tumour biopsies were obtained using an 18-gauge needle, under ultrasound or computed tomography guidance. Only nine minor complications were recorded. Most tumour biopsy samples collected were intended for ancillary molecular studies including protein or gene expression analysis, comparative genomic hybridization array or DNA sequencing. According to the results available, 70% of the biopsy samples met the quality criteria of each study and were suitable for ancillary studies. CONCLUSIONS: In our experience, the majority of the patients accepted skin biopsies as well as tumour biopsies. Sequential tumour and skin biopsies are feasible and safe during early-phase clinical trials, even when patients are exposed to anti-angiogenic agents. The real scientific value of such biopsies for dose selection in phase I trials has yet to be established.

Contrast ultrasonography: necessity of linear data processing for the quantification of tumor vascularization.

PURPOSE: This study is intended to compare the value of uncompressed ultrasonic data, obtained after linear power detection of the ultrasonic radiofrequencies that we call linear data, with usual compressed video data for the quantification of tumor perfusion, particularly for monitoring antivascular therapy. MATERIALS AND METHODS: To form a clinically useful ultrasonic image, the detected power of the received signals (linear data) is compressed in a quasi-logarithmic fashion in order to match the limited dynamic range of the video monitor. The resulting reduced range of signals from an injected contrast agent may limit the sensitivity to changes in the time-intensity curves. Following a theoretical evaluation of the effects of compression on time-intensity curves and as an in vivo example, we measured at different times the effects of an antivascular drug administered to mice bearing melanoma tumors. The mean time-intensity curves within the tumors after bolus injection of a contrast agent were determined using both linear and video data. Linearized data was recovered using the inverse of the true scanner's compression law, which was experimentally determined. Three features were extracted from the time-intensity curves: peak intensity (PI), time to peak intensity (TPI) and area under the curve in the wash-in phase (AUC (wash-in)). When contrast reached its maximum value, the coefficient of variation reflecting the heterogeneity of the intensity of contrast uptake within the tumor, was computed using both data sets. RESULTS: TPI was found to be similar with either data set (r = 0.98, p < 0.05, factor of 1.09). Linear PI and AUC (wash-in) had significantly earlier decreases after drug administration than video data (p = 0.015 and p = 0.03, respectively). The coefficient of variation was significantly lower when using video rather than linear data (p < 10 (-4)). CONCLUSION: In conclusion, the use of linear data is the only mathematically valid methodology for determining a tumor's time-intensity curve and, in practice, it allows earlier demonstration of responses to antivascular drugs.

[Renal cell carcinoma: which criteria to define antiangiogenic treatment failure? A case report]. / Cancer du rein: quels sont les critères pour juger de l'échec d'un traitement anti-angiogénique? A propos d'un cas.

Failure criteria of antiangiogenic treatment that must make envisage a change of molecule are still difficult to define. Concerning the evaluation of the response, if the RECIST criteria seem to be limited, several other tools of evaluation (clinical, radiological or biological tools) can be interesting. It is the case of contrast-enhanced ultra-sonography, but a precise definition of functional parameters should be defined and a standardization of the technique is necessary. Side effects do not translate necessarily a treatment failure. They must be estimated by taking into account the frequency of some of symptoms. Asthenia is noticed in more than 50% of the patients ; it is however necessary to exclude another aetiology, in particular iatrogenic hypothyroidism.

Sunitinib efficacy in the treatment of metastatic skin adnexal carcinomas: report of two patients with hidradenocarcinoma and trichoblastic carcinoma.

BACKGROUND: Adnexal carcinomas are rare and diverse cutaneous tumours. They are locally aggressive and have the potential for distant metastasis. Metastatic adnexal carcinomas are very resistant to conventional chemotherapies. Sunitinib, an oral tyrosine kinase inhibitor, is reportedly effective for the treatment of various solid cancers. Its use in adnexal carcinomas has never been reported. OBSERVATIONS: The first patient had metastatic clear cell hidradenocarcinoma and was stabilized over 8 months with sunitinib, before she relapsed. The second patient had a metastatic malignant hair follicle tumour (trichoblastic carcinoma) and achieved a partial remission with sunitinib, and disease stabilized after 10 months. Dynamic contrast-enhanced ultrasound (DCE-US) performed to evaluate tumour vascularization during treatment depicted a dramatic and early decrease in the tumour blood volume. CONCLUSIONS: Sunitinib was effective in controlling the disease in our two patients. DCE-US using linear raw data may have an early predictive value for tumour response to sunitinib. Further studies involving larger cohorts of patients are warranted in order to confirm the efficacy of sunitinib in these rare tumours.