The Pap smear revisited

Deep Synthesis of Realistic Medical Images: A Novel Tool in Clinical Research and Training

Making clinical decisions based on medical images is fundamentally an exercise in statistical decision-making. This is because in this case, the decision-maker must distinguish between image features that are clinically diagnostic (i.e., signal) from a large amount of non-diagnostic features. (i.e., noise). To perform this task, the decision-maker must have learned the underlying statistical distributions of the signal and noise to begin with. The same is true for machine learning algorithms that perform a given diagnostic task. In order to train and test human experts or expert machine systems in any diagnostic or analytical task, it is advisable to use large sets of images, so as to capture the underlying statistical distributions adequately. Large numbers of images are also useful in clinical and scientific research about the underlying diagnostic process, which remains poorly understood. Unfortunately, it is often difficult to obtain medical images of given specific descriptions in sufficiently large numbers. This represents a significant barrier to progress in the arenas of clinical care, education, and research. Here we describe a novel methodology that helps overcome this barrier. This method leverages the burgeoning technologies of deep learning (DL) and deep synthesis (DS) to synthesize medical images de novo. We provide a proof-of-principle of this approach using mammograms as an illustrative case. During the initial, prerequisite DL phase of the study, we trained a publicly available deep learning neural network (DNN), using open-sourced, radiologically vetted mammograms as labeled examples. During the subsequent DS phase of the study, the fully trained DNN was made to synthesize, de novo, images that capture the image statistics of a given input image. The resulting images indicated that our DNN was able to faithfully capture the image statistics of visually diverse sets of mammograms. We also briefly outline rigorous psychophysical testing methods to measure the extent to which synthesized mammography were sufficiently alike their original counterparts to human experts. These tests reveal that mammography experts fail to distinguish synthesized mammograms from their original counterparts at a statistically significant level, suggesting that the synthesized images were sufficiently realistic. Taken together, these results demonstrate that deep synthesis has the potential to be impactful in all fields in which medical images play a key role, most notably in radiology and pathology.

Screening for cervical neoplasia during pregnancy

OBJECTIVE

To evaluate cytology as a screening method for cervical neoplasia in pregnancy and to compare it with cervicography and the acetic acid test (AAT).

METHODS

In a large antenatal clinic in South Africa, 842 women were screened utilizing cytology, cervicography and the acetic acid test simultaneously. The proportion of positive results of the different tests were compared and the agreement calculated by the kappa statistic.

RESULTS

The mean age of the women was 27 years, and 12.5% smoked. Cytological smears were abnormal (low-grade squamous intra-epithelial lesion and higher degrees of abnormality) in 1.4% of cases, cervicography in 6.3% and the AAT in 14.3% (P = 0.5400). Kappa values were as follows: cytology vs. cervicography 0.01, cytology vs. AAT 0.0 and cervicography vs. AAT 0.2.

CONCLUSIONS

As a result of cytology's rather low yield and the small measure of agreement between the tests, cytology should be supplemented by an additional screening test in pregnancy.

Cytological early detection of cervical carcinoma: possibilities and limitations. Analysis of failures

Gynecological cytology has been intensified in the district of Schwerin since 1968. The incidence of cervical carcinoma decreased from 38.9 per 100,000 women in 1969 to 19.8 in 1991 (50% of the initial figure). The mortality decreased from 25.1 to as little as 7.0 (28% of the initial figure). Because invasive cervical carcinoma is theoretically avoidable, a total of 577 new cases observed from 1980 to 1988 were investigated in respect of participation in screening and reasons for the failures. 317 (55%) of these women did not participate in the examination more than five years before diagnosis of invasive cancer. 260 (45%) of these women were examined by gynecologists once, several times (87 (67%)) or annually (173 (33%)) during a five year period prior to the diagnosis of cancer. An independent review of the smears explained the highest number of negative smears by nonobtaining of representative material and only 19% as a ministerpretation of the cytological finding. The gynecologists failed to: take smears (36 women); carry out a cytological control or histological verification (48); obtain adequate smears (116); consider clinical symptoms (22).

Increased activity of 6-phosphogluconate dehydrogenase and glucose-6-phosphate dehydrogenase in purified cell suspensions and single cells from the uterine cervix in cervical intraepithelial neoplasia

The activities of 6-phosphogluconate dehydrogenase and glucose-6-phosphate dehydrogenase have been measured in squamous epithelial cells of the uterine cervix from normal patients and cases of cervical intraepithelial neoplasia (CIN). A biochemical cycling method, which uses only simple equipment and is suited to routine use and to automation, was applied to cells separated by gradient centrifugation. In addition, cells were examined cytochemically, and the intensity of staining in the cytoplasm of single whole cells was measured using computerised microcytospectrophotometry. Twenty per cent of cells in samples from normal patients (n=61) showed staining intensities above an extinction of 0.15 at 540 nm, compared to 71% of cases of CIN 1 (n=14), 91% of cases of CIN 2 (n=11) and 67% of cases of CIN 3 (n=15). The cytochemical data do not allow definitive distinctions to be made between different grades of CIN whereas the biochemical assay applied to cell lysates shows convincing differences between normal samples and cases of CIN. There are no false negatives for CIN 3 (n=14) and CIN 2 (n=10) and 11% false negatives for CIN 1 (n=9) and 14% of false positives for normal cases (n=21). The results of this preliminary study with reference to automation are discussed [corrected].

Post-irradiation cytology of cervical cancer patients

The accuracy of cervicovaginal cytology following radiotherapy for cervical cancer is compromised by the anatomical and tissue changes resulting from irradiation. Collection of representative samples may be more difficult, and benign radiation changes, post-irradiation dysplasia, and the frequent occurrence of repair cells and active stromal cells in post-irradiation smears may cause diagnostic problems. Nevertheless, cytology is a valuable tool for the detection of locally recurrent cervical cancer. It is simple and economical to perform at the time of clinical follow-up examination, and may detect occult tumour recurrence. Awareness of the cellular changes resulting from irradiation, and the varied composition of post-irradiation smears may lead to more accurate interpretation of the cytological findings.

The cytological detection of persistent cervical intraepithelial neoplasia after local ablative treatment: a comparison of sampling devices

OBJECTIVE

To determine whether the cytological detection of persistent cervical intraepithelial neoplasia (CIN) after local ablative treatment is improved by the use of sampling devices other than the Ayre's spatula.

DESIGN

A randomized controlled study.

SETTING

Lothian Area Colposcopy Clinic.

SUBJECTS

856 patients who had received local therapy (CO2 laser or cold coagulation) for CIN II or III between 9 and 30 months earlier.

INTERVENTION

Each patient had three consecutive cervical smears taken, one with the Ayre's spatula, one with either the Aylesbury, the Rocket or the Multispatula device, and finally one with the Cytobrush. The allocation of which spatula and the order of the first two was randomized. Each patient had a colposcopic examination immediately after the smears were taken.

MAIN OUTCOME MEASURES

A comparison of the detection of histologically proven persistent CIN by the Ayre's spatula with the detection of persistent disease by alternative sampling devices.

RESULTS

Of the 856 patients 130 had histologically proven persistent CIN. Another 98 had suspicious findings on colposcopy but punch biopsies reported as histologically normal. Of the remaining patients with normal colposcopy 130 were randomly selected to form a control group. The cervical smears from these 358 women were reported. Significantly fewer Ayre's samples contained endocervical cells than Aylesbury samples (47% vs 59%, difference 12%; 95% CI 3%-21%; P less than 0.001), Rocket samples (47% vs 67%; difference 20%, 95% CI; 12%-32%; P less than 0.001) or Multispatula samples (47% vs 76%; difference 29%, 95% CI 19-38%; P less than 0.001). When punch biopsies contained CIN, dyskaryotic cells were seen in 10% of Ayre's samples, 4.3% of Aylesbury samples, 8.3% of Rocket samples, and in no smear taken with the Multispatula. Obtaining a third smear with the Cytobrush did not substantially improve the detection rate of dyskaryosis. Neither the order of use of the spatulas, the form of initial treatment nor the size of the transformation zone had any apparent effect on the cytological detection of persistent CIN.

CONCLUSIONS

We recommend that surveillance of patients who have received local ablative therapy for CIN should be by both cytology and colposcopy, and that cytological samples should be obtained using the Ayre's spatula.

Human papillomavirus infection and cervical intraepithelial neoplasia in women with renal allografts

An increased prevalence of cervical cancer has been observed in immunosuppressed women, but controlled studies are rare. Biopsy specimens from 49 women with renal allografts and 69 non-immunosuppressed controls (with no history of cervical intraepithelial neoplasia, vulval warts, or abnormal results of cervical smear tests) were assessed for colposcopic appearance, cytological and histological diagnosis, and the presence of human papillomavirus types 6/11 and 16/18 DNA sequences. At colposcopy 26 (53%) of the women with allografts had cervical abnormalities compared with 20 (29%) of the controls. The prevalence of cervical intraepithelial neoplasia was significantly higher in the women with allografts (24 (49%) compared with 7 (10%]. The overall rate of detection of human papillomavirus DNA did not differ significantly between the two groups. There was however, a significant difference in the rate of detection of human papillomavirus type 16/18 DNA (27% in the women with allografts and 6% in the controls). These data confirm that pathological and virological changes affecting the cervix are significantly increased in immunosuppressed women and emphasise the need for regular colposcopic examination.