Artificial intelligence and breast screening: French Radiology Community position paper.

The objective of this article was to evaluate the evidence currently available about the clinical value of artificial intelligence (AI) in breast imaging. Nine experts from the disciplines involved in breast disease management - including physicists and radiologists - convened a meeting on June 3, 2019 to discuss the evidence for the use of this technology in plenary and focused sessions. Prior to the meeting, the group performed a literature review on predefined topics. This paper presents the consensus reached by this working group on recommendations for the future use of AI in breast screening and related research topics.

Changes by age in breast cancer incidence, mammography screening and hormone therapy use in France from 2000 to 2006.

BACKGROUND: In 2003, US breast cancer incidence rates fell. Recent French data reveal also a decline in 2005-2006. This study aims to present the trends in breast cancer incidence by age and to identify the respective impact of mammography screening and use of hormone replacement therapy (HRT) in the French context. METHODS: Breast cancer incidence rates were calculated from the new cases of breast cancer among affiliates of the general scheme of the French National Health Fund between 2000 and 2006. Data concerning HRT and mammograms were extracted from the reimbursement databanks of the National Health Fund and from the National Screening Programme. RESULTS: Breast cancer incidence decreased between 2003 and 2006 only for women aged 50 or above. The strongest declines were observed among the 55-59 and 60-64-year-old groups (12.9 and 7.7%, respectively). We observed a slight decline in the age groups of 50-54 and 65-69 (0.7 and 2.1%, respectively). Volumes of mammograms increased continuously between 2000 and 2006 from 1,600,000 to 3,470,000 for women aged 50-74 years old. In 2004, the National Screening Programme achieved complete geographic coverage. At the same time, the number of HRT users has dropped by 62% between 2001 and 2006. We observed the highest prevalence of HRT and the highest decrease in breast cancer incidence rates in the age group of 55-59. CONCLUSIONS: The recent reduction in breast cancer incidence in France for women aged 50 years or above, in 2005-2006, was accompanied by a substantial reduction in HRT prescriptions after 2002 for all age groups. The drop in HRT parallels the drop in breast cancer incidence for the women between the ages of 55-59 and 60-64. The high-level of development of screening in France during the same period could not account for the reduction in breast cancer incidence.

[Breast cancer screening: are results of French and international programmes comparable?]. / Dépistage organisé des Cancers du Sein: peut-on comparer les résultats du programme français aux résultats internationaux?

Full coverage of the national breast cancer screening programme was obtained in 2004. The new protocol was published in 2001. It changed the organisation implemented since 1994 for a better fit with the French decentralized health care system. Consequences on health results were analysed explaining related medical and sociological factors. The main change was the rate of positive mammograms which differed from the recall rate given in the European recommendations. With the new protocol the rate of assessments more than doubled: 6% in year 2000 and 14% in 2003. International comparisons are difficult since each country has a different organisation. Moreover the coexistence in France of organized screening and opportunistic screening causes a complex issue for evaluation of impact indicators of this population based programme.

Comparison of early performance indicators for screening projects within the European Breast Cancer Network: 1989-2000.

In 1989 the European Breast Cancer Network (EBCN) was established by the first pilot projects for breast cancer screening, co-funded by the Europe Against Cancer programme. We report early performance indicators for these EBCN projects while taking into account their organizational setting. Out of 17 projects in the network, 10 projects from six European countries contributed aggregated data on number of invitations, screening examinations, and breast cancers detected over the period 1989-2000. Results were summarized separately for projects in centralized versus decentralized health care environments. The European Guidelines for quality assurance in mammography screening provided reference values for the performance indicators. The most prominent finding in this study was the higher participation rate in centralized versus decentralized projects (average participation in 1998: 74 versus 33%; P<0.001), whereas the invitation system and screening policy in these projects were similar. Detection rates and characteristics of cancers detected at initial and subsequent screening examinations showed no significant differences between centralized and decentralized projects. Even though early performance indicators for centralized versus decentralized projects were similar, the impact of breast screening on mortality from this disease at the population level will differ since the decentralized projects reach only part of the target population.

Does hormone replacement therapy increase the frequency of breast atypical hyperplasia in postmenopausal women? Results from the Bouches du Rhone district screening campaign.

It is thought that the risk of atypical hyperplasia (AH) increases with age, particularly among postmenopausal women. Three hypotheses were investigated to try to explain this phenomena: use of hormone replacement therapy (HRT), increased breast cancer screening and improvements in radiological quality. Data were collected from the Bouches du Rhône breast cancer screening programme database and from the pathological registry of all women operated on for breast diseases in the district. The AH incidence rate was studied using a Poisson regression analysis. The change in the profile of breast diseases was explored through studying changes in the proportion of AH among benign lesions and malignant diseases. The AH incidence rate significantly increased over time (13.6% per year). The proportion of AH among the benign diseases increased with time and was significantly higher for HRT users (Odds Ratio (OR)=2.05; 95% Confidence Interval (CI): 1.43-2.93). While AH decreased with age among HRT non-users, it increased among users as a proportion of both benign and malignant lesions. The AH incidence rate significantly increased among pre- and postmenopausal women. Our study suggests that this increase is partly explained by the incidental discovery of these lesions by mammography and partly by a real increase of the disease among HRT users.

Does a better grade of tumour occurring in women under hormone replacement therapy compensate for their lower probability of detection by screening mammography.

OBJECTIVE: To compare the prognostic factor of breast cancer survival between breast cancer diagnosed in subjects receiving hormone replacement therapy (HRT) before diagnosis to those without such a therapy. SUBJECTS AND METHODS: All breast cancers diagnosed between 1993 and 2000 within the breast cancer screening programme in Bouches du Rhône (France) were analysed for size, node status, and grade according to use, or not, of HRT. Univariate and multivariate analyses were carried out taking into account age, density of the breast, and mode of detection. RESULTS: The breast tumours diagnosed among HRT users had a lower grade whatever the mode of detection. The proportion of node positive tumours was identical in the two groups after adjustment for age. The smaller size of the tumours among HRT users is partly explained by the lower grade of these tumours Conclusion: Although tumours occurring in HRT users have a lower chance of being detected by screening, their prognostic factors, especially the grade of the tumour, are better than in non-users. More work is needed to find which part of this advantage is attributable to better surveillance of women treated with HRT

[Digital mammography and computer assisted diagnosis]. / La mammographie numérique: technique, applications et apport de l'aide informatisée au diagnostic.

Until recently, the film has remained the only medium of information in mammography. The film is used to record, to exploit, to store and to transmit the image. Computerization of images allows to dissociate and to optimize these different functions. Among classical factors of image quality (spatial resolution, contrast, noise), new factors should be added like the detective quantum efficiency and the conversion factor. Radioluminescent screens, then digital sensors for breast stereotactic imaging have been marketed. Manufacturers are now testing full field digital mammographs. Digital imaging allows many applications (computed-aided diagnostic, 3D imaging.) and permits the easy transfer of images for diagnosis and teaching. Three parts are presented in this chapter. The first one describes the different imaging modalities and gives a reminder of the different elements related to image quality. The second one is related to the practical aspects of full field mammography, the reading of mammograms on a review station, ergonomy in full field mammography and to possible changes for screening mammography. The third part is devoted to computed aided diagnosis and its possible application in screening.

[The cost of breast cancer screening in France]. / Le coût du dépistage du cancer du sein en France.

PURPOSE: This paper presents the cost of two decentralised breast cancer screening programmes in France, in the Bas-Rhin and the Bouches du Rhône districts. Materials and methods. The costs directly related to running the two screening programmes were collected for the time period 1990-1997. Only direct costs of each programme, excluding assessment and treatment costs, are included in the cost analysis. Costs are presented per screening period, per programme attender and per cancer detected by screening. RESULTS: The screening programme costs an average of 397 francs per woman screened and 90,828 francs per cancer detected in the Bas-Rhin, and 379 francs and 76,159 francs, respectively, in the Bouches du Rhône. CONCLUSIONS: These costs are high compared to those of existing centralised programmes. Further research is needed to investigate means of ensuring the optimal efficiency of these programmes. The results of this study may help guide future decisions on the further development of breast cancer screening in France.

[Anthropological studies of the representations of menopause in a population of women invited to a mass screening for breast cancer in Bouches-du-Rhône and Charente]. / Approche anthropologique des représentations de la ménopause dans une population de femmes invitées au dépistage de masse organisé des cancers du sein--Bouches-du-Rhône et Charente.

This is an anthropological study of a target population (breast screening women) in the Bouches-du-Rhône and Charente regions of France. The occurrence of menopause is comparable in these two departments and depends on providing medical care which leads to breast screening. Menopause is natural for these women and is considered as a sign of ageing. Psychosomatic symptoms vary with sociocultural groups. Hormonal replacement therapy furthers breast screening. In the popular imagination, there is a deficit between nature (non-HRT) and culture (HRT). They take estrogen in the form of soy to offset this inadequacy, which creates a new cultural syncretism.

Hormone replacement therapy and screening mammography: analysis of the results in the Bouches du Rhône programme.

OBJECTIVE: To evaluate the effectiveness of a mass screening programme for breast cancer in a French population where hormone replacement therapy (HRT) is common and where mammography is prescribed outside the programme for asymptomatic women. METHODS: From 1993 to 1996 inclusive, 41,062 women underwent a first test and 48,275 a second or third test in the Bouches du Rhône programme. Their HRT status was ascertained at the time of the test. False positive and false negative tests were identified at one year follow up. The incidence of interval cancers was estimated up to three years after the screening test. RESULTS: The odds of being detected at screening rather than as an interval cancer within one year of the test was five times greater among non-users of HRT than among users (odds ratio (OR) 5.14 (confidence interval 2.5 to 11.8)). This high reduction in sensitivity among users (71% v 92%) was associated with a very small reduction in specificity at the incident screen only. The incidence of interval cancers among HRT users was 3.5 times that of non-users within the first year after the test and 1.7 times during the following two years. CONCLUSIONS: When the early results of a programme have been used to measure its effectiveness they should be reassessed in populations where HRT is in widespread use. As interval cancers had better prognostic factors in HRT users than in non-users, the efficacy of a screening programme in such populations should be the subject of further studies.

[Digital image processing of screening mammographies: application to masses]. / Traitement des mammographies de dépistage par ordinateur: application aux opacités.

A method based on image processing of tumoral neighbourhood is described for studying masses. By using polar and pseudopolar representations for the tumor and its neighbourhood, it is possible to detect divergent structures around the lesion and to evidence fuzzy areas in the boundary of the tumor, which constitute two significant features of breast cancer. The degree of spiculation is deduced from a shape parameter that characterizes the irregularity of the boundary; the radial component of the gradient on the boundary provides a measurement of fuzzy appearance.

[French screening program of breast cancer. Results from 5 districts (1989-1994)]. / Programme français de dépistage du cancer du sein. Résultats de cinq départements (1989-1994.

Randomized clinical trials have demonstrated that screening by mammography can reduce the breast cancer mortality rate by 30 to 50% in women aged 50 to 69 years. In France, pilot screening programmes were begun in 1989, and a national programme is being implemented since 1994. After 8 years of experience, the present study aims to assess the effectiveness of the French decentralized model of screening. This study presents data obtained by the National Steering Committee from the reports of five district programmes. Efficacy indicators from 1989 to 1994 are presented and compared with European targets. Data are presented by screening round. The prevalent round concerns 260,226 women screened, the subsequent incident rounds concern 119,761 women. Attendance rates at 5 years range from 24.2% to 50.9%. The mean recall rate was 7.8% in the first and 4.5% in the second round 5.4 and 2.7 cancers were detected per 1,000 women screened during the first and second round, respectively. and 30% of invasive cancers were of 10 mm or less and 70% had no nodal involvement. The interval cancer rate was 0.55% one year after screening and 1.05% two years after screening. Results from our decentralized screening system are variable yet satisfactory overall, namely thanks to the progressive implementation of quality assurance. However, attendance rates remain low, whereas actual mammographic coverage is 60% in France. A generalized screening programme can only be foreseen in France if its implementation is very gradual and if the quality of screening is high. The improvement of radiological quality remains the primary objective of the National Steering Committee however this implies training of all personnel. In the long run, organized screening should prevail over spontaneous screening. It is unlikely that we will achieve the target reduction in mortality rate in our country.

[Opportunistic screening of breast cancer in France]. / Le dépistage spontané du cancer du sein en France.

In France, one fears that opportunistic screening levels have risen in parallel with the development of district screening programmes. However the absence of a specific nomenclature for mammography renders its evaluation difficult. This study aims to quantify the importance of opportunistic screening in 2 districts with existing screening programmes. Mammograms done outside of the screening programmes were identified using Sickness Fund data. Reasons for prescription were identified through a survey of radiologists. These data were compared to programme attendance rates. In 1995, 9% of women aged 40 years or more had a mammogram outside of the screening programme. This rate is higher among women under the age of 55 when compared to older women. Opportunistic screening accounted for 27% of mammograms outside the screening programme, yet this rate reached 58% if one includes mammograms in asymptomatic women with personal or family risk factors. The data presented in this study confirm the importance of opportunistic screening in France, namely in peri-menopausal women. Screening programmes allow for the continuous evaluation of the quality and effectiveness of screening. The coexistance of opportunistic and organized screening should not be permitted to continue. The results of ongoing cost-effectiveness studies will allow to determine the relative cost-effectiveness of organized screening in France.

A comparison of the performance and impact of breast cancer screening programmes in East Anglia, U.K. and Bouches du Rhône, France.

The purpose of this study was to compare results from the first screening round of two breast cancer screening programmes of similar design implemented in different health care settings. The East Anglian programme is part of the U.K. National Health Service Breast Screening Programme, which is a centralised programme with a limited number of dedicated screening units. The Bouches du Rhône programme is one of 13 French programmes based on a decentralised model using existing radiology clinics. Compliance and cancer detection rates were lower in the Bouches du Rhône programme. Detection rates for small invasive cancers (< or = 10 mm) were similar in the two programmes, although larger cancers (> or = 20 mm) were detected in the Bouches du Rhône programme. Significantly, the shift towards more favourable distribution of prognostic characteristics associated with screen-detected breast cancers compared with those arising outside the programme is less marked in the Bouches du Rhône programme. This is probably due to the more favourable underlying disease status in the district resulting from a long history of breast awareness and spontaneous mammography.

[Modalities of reading of detection mammographies of the programme in the Bouches-du-Rhône. Results and costs 1990-1995]. / Modalités de lecture des mammographies de dépistage du programme des Bouches-du-Rhône. Résultats et coûts 1990-1995.

In the Bouches du Rhône breast cancer screening programme, mammograms are read by two separate radiologists of different levels of training. All women with at least one positive reading are recalled for further assessment. During the first round, 3,477 of the 95,967 screenees were recalled by the first reader and 2,321 by the expert reader. The expert increased cancer detection by 15% and 45% of cancers detected by the expert were smaller than 11 mm. The marginal cost of double reading was 21,838 Francs per additional cancer detected. Double reading thus allowed for the detection of cancers of good prognosis which would have gone undetected by a single reader. Its cost seems justified by its impact on the effectiveness of the screening programme.

Detection of five different human papillomavirus types in cervical lesions by in situ hybridization. A study of 110 cases using sulfonated probes.

Several findings suggest an etiologic relationship between genital tract squamous cell carcinoma and certain types of human papillomavirus (HPV). Detection of these HPV types in cervical lesions considered as preneoplastic states (ie, cervical intraepithelial neoplasia or CIN) is extremely important but difficult because the morphology of these states is highly heterogeneous and clinical course is rarely predictable. In situ hybridization (ISH) is the only technique allowing correlation between HPV type and tissue or cell morphology. In this report, 110 biopsy specimens from uterine cervix lesions were studied: 66 CIN, 10 invasive carcinoma, 28 metaplasia, and six condyloma acuminata. A new ISH technique based on direct modification of DNA probes by sulfonation was used. The hybridized DNA was revealed first by a specific monoclonal antibody against sulfonated DNA, and then by an alkaline phosphatase system. In order to determine the sensitivity level of this method, 14 biopsy specimens were also submitted to Southern blot hybridization. Five probes were used separately (HPV 6, 11, 16, 18, and 33) for each biopsy specimen. Results of ISH were correlated with morphologic criteria such as number of koilocytes and mitoses. Oncogenic HPV was found exclusively in CIN. The number of labeled cells varied with CIN grade. These data suggest that, whatever the grade, CIN represents a unique preneoplastic process, and that HPV replication depends on the squamous maturation of the pathologic epithelium.

[Epidemiologic survey on viral lesions and CIS of the cervix: multicenter retrospective study in public and private centers]. / Enquête épidémiologique sur les lésions virales et CIN du col: étude multicentrique rétrospective dans des centres publiques et privés.

Fifteen pathologists working in various laboratories contributed to a survey designed to evaluate the frequency of viral lesions and carcinoma in situ of the cervix. This study, based on almost 400,000 cervico-vaginal smears demonstrated that the lesions were fell into a relatively narrow range (between 0.2% and 3% of low-grade lesions, between 0.2 and 2% of high-grade lesions, total at risk lesions between 0.5 and 5.2%). The results were stable over time within the one center. Sub-typing of human papilloma viruses provides a more informative classification of the low-grade lesions in which the morphological frontier between carcinoma in situ and condyloma is often blurred and in which the detection of an oncogenic virus lead to more rapid treatment.