On-the-fly point annotation for fast medical video labeling

PURPOSE

In medical research, deep learning models rely on high-quality annotated data, a process often laborious and time-consuming. This is particularly true for detection tasks where bounding box annotations are required. The need to adjust two corners makes the process inherently frame-by-frame. Given the scarcity of experts' time, efficient annotation methods suitable for clinicians are needed.

METHODS

We propose an on-the-fly method for live video annotation to enhance the annotation efficiency. In this approach, a continuous single-point annotation is maintained by keeping the cursor on the object in a live video, mitigating the need for tedious pausing and repetitive navigation inherent in traditional annotation methods. This novel annotation paradigm inherits the point annotation's ability to generate pseudo-labels using a point-to-box teacher model. We empirically evaluate this approach by developing a dataset and comparing on-the-fly annotation time against traditional annotation method.

RESULTS

Using our method, annotation speed was 3.2 × faster than the traditional annotation technique. We achieved a mean improvement of 6.51 ± 0.98 AP@50 over conventional method at equivalent annotation budgets on the developed dataset.

CONCLUSION

Without bells and whistles, our approach offers a significant speed-up in annotation tasks. It can be easily implemented on any annotation platform to accelerate the integration of deep learning in video-based medical research.

Investigating genotype-to-phenotype correlation in CHARGE syndrome by deep phenotyping and multiparametric clustering

CHARGE syndrome, due to CHD7 pathogenic variations, is an autosomal dominant disorder characterized by a large spectrum of severity. Despite the great number of variations reported, no clear genotype-to-phenotype correlation has been reported. Unsupervised machine learning and clustering was undertaken using a retrospective cohort of 42 patients, after deep radiologic and clinical phenotyping, to establish genotype-phenotype correlation for CHD7-related CHARGE syndrome. It resulted in three clusters showing phenotypes of different severities. While no clear genotype-phenotype correlation appeared within the first two clusters, a single patient was outlying the cohort data (cluster 3) with the most atypical phenotype and the most distal frameshift variant in the gene. We added two other patients with similar distal pathogenic variants and observed a tendency toward mild and/or atypical phenotypes. We hypothesized that this finding could potentially be related to escaping nonsense mediated RNA decay, but found no evidence of such decay in vivo for any of the CHD7 pathogenic variation tested. This indicates that this milder phenotype may rather result from the production of a protein retaining all functional domains.

CT-based Bosniak classification of cystic renal lesions: is version 2019 an improvement on version 2005?

OBJECTIVE

To compare the diagnostic performance and inter-reader agreement of the CT-based v2019 versus v2005 Bosniak classification systems for risk stratification of cystic renal lesions (CRL).

METHODS

This retrospective study included adult patients with CRL identified on CT scan between 2005 and 2018. The reference standard was histopathology or a minimum 4-year imaging follow-up. The studies were reviewed independently by five readers (three senior, two junior), blinded to pathology results and imaging follow-up, who assigned Bosniak categories based on the 2005 and 2019 versions. Diagnostic performance of v2005 and v2019 Bosniak classifications for distinguishing benign from malignant lesions was calculated by dichotomizing CRL into the potential for ablative therapy (III-IV) or conservative management (I-IIF). Inter-reader agreement was calculated using Light's Kappa.

RESULTS

One hundred thirty-nine patients with 149 CRL (33 malignant) were included. v2005 and v2019 Bosniak classifications achieved similar diagnostic performance with a sensitivity of 91% vs 91% and a specificity of 89% vs 88%, respectively. Inter-reader agreement for overall Bosniak category assignment was substantial for v2005 (κ = 0.78) and v2019 (κ = 0.75) between senior readers but decreased for v2019 when the Bosniak classification was dichotomized to conservative management (I-IIF) or ablative therapy (III-IV) (0.80 vs 0.71, respectively). For v2019, wall thickness was the morphological feature with the poorest inter-reader agreement (κ = 0.43 and 0.18 for senior and junior readers, respectively).

CONCLUSION

No significant improvement in diagnostic performance and inter-reader agreement was shown between v2005 and v2019. The observed decrease in inter-reader agreement in v2019 when dichotomized according to management strategy may reflect the more stringent morphological criteria.

KEY POINTS

• Versions 2005 and 2019 Bosniak classifications achieved similar diagnostic performance, but the specificity of higher risk categories (III and IV) was not increased while one malignant lesion was downgraded to v2019 Bosniak category II (i.e., not subjected to further follow-up). • Inter-reader agreement was similar between v2005 and v2019 but moderately decreased for v2019 when the Bosniak classification was dichotomized according to the potential need for ablative therapies (I-II-IIF vs III-IV).

Conventional and artificial intelligence-based imaging for biomarker discovery in chronic liver disease

Chronic liver diseases, resulting from chronic injuries of various causes, lead to cirrhosis with life-threatening complications including liver failure, portal hypertension, hepatocellular carcinoma. A key unmet medical need is robust non-invasive biomarkers to predict patient outcome, stratify patients for risk of disease progression and monitor response to emerging therapies. Quantitative imaging biomarkers have already been developed, for instance, liver elastography for staging fibrosis or proton density fat fraction on magnetic resonance imaging for liver steatosis. Yet, major improvements, in the field of image acquisition and analysis, are still required to be able to accurately characterize the liver parenchyma, monitor its changes and predict any pejorative evolution across disease progression. Artificial intelligence has the potential to augment the exploitation of massive multi-parametric data to extract valuable information and achieve precision medicine. Machine learning algorithms have been developed to assess non-invasively certain histological characteristics of chronic liver diseases, including fibrosis and steatosis. Although still at an early stage of development, artificial intelligence-based imaging biomarkers provide novel opportunities to predict the risk of progression from early-stage chronic liver diseases toward cirrhosis-related complications, with the ultimate perspective of precision medicine. This review provides an overview of emerging quantitative imaging techniques and the application of artificial intelligence for biomarker discovery in chronic liver disease.

Comparison of Transient Elastography, ShearWave Elastography, Magnetic Resonance Elastography and FibroTest as routine diagnostic markers for assessing liver fibrosis in children with Cystic Fibrosis

BACKGROUND AND OBJECTIVE

Reliable markers are needed for early diagnosis and follow-up of liver disease in Cystic Fibrosis (CF). The objective was to evaluate the diagnostic performance of Transient Elastography (TE), Real-Time ShearWave Ultrasound Elastography (SWE), Magnetic Resonance Elastography (MRE) and the FibroTest as markers of Cystic Fibrosis Liver Disease (CFLD).

METHODS

A monocentric prospective cross-modality comparison study was proposed to all children (6 to 18 years of age) attending the CF center. Based on liver ultrasound findings, participants were classified into 3 groups: multinodular liver or portal hypertension (Nodular US/PH, advanced CFLD), heterogeneous increased echogenicity (Heterogeneous US, CFLD) or neither (Normal/Homogeneous US, no CFLD). The 4 tests were performed on the same day. The primary outcome was the FibroTest value and liver stiffness measurements (LSM).

RESULTS

55 participants (mean age 12.6 ± 3.3 years; 25 girls) were included between 2015 and 2018: 23 in group Nodular US/PH, 8 in group Heterogeneous US and 24 in group Normal/Homogeneous US (including 4 with steatosis). LSM on TE, SWE and MRE were higher in participants with CFLD (groups Nodular US/PH and Heterogeneous US) compared to others (group Normal/Homogeneous US) (p<0.01), while FibroTest values did not differ (p = 0.09). The optimal cut-off values for predicting CFLD on TE, SWE and MRE were 8.7 (AUC=0.83, Se=0.71, Sp=0.96), 7.8 (AUC=0.85, Se=0.73, Sp=0.96) and 4.15 kPa (AUC=0.68, Se=0.73, Sp=0.64), respectively. LSM predicted the occurrence of major liver-related events at 3 years. TE and SWE were highly correlated (Spearman's ρ=0.9) and concordant in identifying advanced CFLD (Cohen's κ=0.84) while MRE was moderately correlated and concordant with TE (ρ=0.41; κ=36) and SWE (ρ=0.5; κ=0.50).

CONCLUSION

This study demonstrated excellent diagnostic performance of TE, SWE and MRE for the diagnosis of CFLD.

Lumacaftor-ivacaftor effects on cystic fibrosis-related liver involvement in adolescents with homozygous F508 del-CFTR

BACKGROUND

The effects of lumacaftor-ivacaftor on cystic fibrosis transmembrane conductance regulator (CFTR)-associated liver disease remain unclear. The objective of the study was to describe the effect of this treatment on features of liver involvement in a cystic fibrosis (CF) adolescent population homozygous for F508del.

METHODS

Clinical characteristics, liver blood tests, abdominal ultrasonography (US), and pancreas and liver proton density fat fraction (PDFF) by magnetic resonance imaging, were obtained at treatment initiation and at 12 months for all patients. Biomarkers of CFTR activity (sweat chloride test, nasal potential difference, and intestinal current measurement) were assessed at initiation and at 6 months therapy.

RESULTS

Of the 37 patients who started ivacaftor/lumacaftor treatment, 28 were eligible for analysis. In this group, before treatment initiation, 4 patients were diagnosed with multinodular liver and portal hypertension, 19 with other forms of CF liver involvement, and 5 with no signs of liver involvement. During treatment, no hepatic adverse reactions were documented, and no patient developed liver failure. Serum levels of alanine aminotransferase (ALT), aspartate aminotransferase (AST) and gammaglutamyl transferase (GGT) decreased significantly following initiation of lumacaftor-ivacaftor, and remained so after 12 months treatment. This was not correlated with changes in clinical status, liver and pancreas US and PDFF, fecal elastase, or lumacaftor-ivacaftor serum levels. The most "responsive" patients demonstrated a significant increase in biomarkers of CFTR activity.

CONCLUSIONS

These results may suggest a potential beneficial effect of CFTR modulators on CF liver disease and warrant further investigation in larger, prospective studies.

Artificial intelligence-based radiomics models in endometrial cancer: A systematic review

BACKGROUND

Radiological preoperative assessment of endometrial cancer (EC) is in some cases not precise enough and its performances improvement could lead to a clinical benefit. Radiomics is a recent field of application of artificial intelligence (AI) in radiology.

AIMS

To investigate the contribution of radiomics on the radiological preoperative assessment of patients with EC; and to establish a simple and reproducible AI Quality Score applicable to Machine Learning and Deep Learning studies.

METHODS

We conducted a systematic review of current literature including original articles that studied EC through imaging-based AI techniques. Then, we developed a novel Simplified and Reproducible AI Quality score (SRQS) based on 10 items which ranged to 0 to 20 points in total which focused on clinical relevance, data collection, model design and statistical analysis. SRQS cut-off was defined at 10/20.

RESULTS

We included 17 articles which studied different radiological parameters such as deep myometrial invasion, lympho-vascular space invasion, lymph nodes involvement, etc. One article was prospective, and the others were retrospective. The predominant technique was magnetic resonance imaging. Two studies developed Deep Learning models, while the others machine learning ones. We evaluated each article with SRQS by 2 independent readers. Finally, we kept only 7 high-quality articles with clinical impact. SRQS was highly reproducible (Kappa = 0.95 IC 95% [0.907-0.988]).

CONCLUSION

There is currently insufficient evidence on the benefit of radiomics in EC. Nevertheless, this field is promising for future clinical practice. Quality should be a priority when developing these new technologies.

Multimodality Imaging and Artificial Intelligence for Tumor Characterization: Current Status and Future Perspective

Research in medical imaging has yet to do to achieve precision oncology. Over the past 30 years, only the simplest imaging biomarkers (RECIST, SUV,…) have become widespread clinical tools. This may be due to our inability to accurately characterize tumors and monitor intratumoral changes in imaging. Artificial intelligence, through machine learning and deep learning, opens a new path in medical research because it can bring together a large amount of heterogeneous data into the same analysis to reach a single outcome. Supervised or unsupervised learning may lead to new paradigms by identifying unrevealed structural patterns across data. Deep learning will provide human-free, undefined upstream, reproducible, and automated quantitative imaging biomarkers. Since tumor phenotype is driven by its genotype and thus indirectly defines tumoral progression, tumor characterization using machine learning and deep learning algorithms will allow us to monitor molecular expression noninvasively, anticipate therapeutic failure, and lead therapeutic management. To follow this path, quality standards have to be set: standardization of imaging acquisition as it has been done in the field of biology, transparency of the model development as it should be reproducible by different institutions, validation, and testing through a high-quality process using large and complex open databases and better interpretability of these algorithms.

Spontaneous gas in a retroperitoneal mass: check the testis!

Testicular germ cell tumor is the most common cancer in 20-to 35-years-old men. There are known risk factors such as undescended testicle(s) and history of testicular cancer. Most lesions are germ cell tumors with two main subtypes: seminomas and non-seminomatous germ cell tumors.

Interactions of doctors with the pharmaceutical industry

OBJECTIVE

To assess the opinions and practice patterns of obstetrician-gynaecologists on acceptance and use of free drug samples and other incentive items from pharmaceutical representatives.

METHODS

A questionnaire was mailed in March 2003 to 397 members of the American College of Obstetricians and Gynecologists who participate in the Collaborative Ambulatory Research Network.

RESULTS

The response rate was 55%. Most respondents thought it proper to accept drug samples (92%), an informational lunch (77%), an anatomical model (75%) or a well-paid consultantship (53%) from pharmaceutical representatives. A third (33%) of the respondents thought that their own decision to prescribe a drug would probably be influenced by accepting drug samples. Respondents were more likely to think the average doctor's prescribing would be influenced by acceptance of the items than theirs would be (p<0.002). Respondents who distributed drug samples to patients indicated doing so because of patients' financial need (94%) and for their convenience (76%) and less so as a result of knowledge of the efficacy of the sample product (63%). A third (34%) of respondents agreed that interactions with industry should be more strictly regulated.

CONCLUSION

Obstetrician-gynaecologists largely indicated that they would act in accordance with what they think is proper regarding accepting incentive items from pharmaceutical representatives. Although accepting free drug samples was considered to be appropriate more often than any other item, samples were most commonly judged to be influential on prescribing practices. The widely accepted practice of receiving and distributing free drug samples needs to be examined more carefully.