All cancers must be microscopically diagnosed and classified before they can be treated. This can be performed subjectively on cells or tissues under a light microscope cytologically or histologically by skilled personnel, mostly pathologists. Well-educated cytotechnicians may assist in prescreening cytological specimens. While the bioptic acquisition of tissue needs scalpels or biopsy needles and local anesthesia, that of cells needs neither. They can be obtained noninvasively either from body fluids (cerebrospinal fluid, serous effusions, sputa, bronchial secretions, or urine) or by brushing mucosal surfaces (conjunctival, oral, pharyngeal, laryngeal, bronchial, bile ducts, cervical, or perianal). Using fine-needle aspiration, biopsy cells can also be obtained from inner organs (thyroid, salivary glands, lungs, lymph nodes, liver, pancreas, kidneys, or prostate). The compliance of patients for a cytological clarification of a given suspicious lesion (e.g., an oral leukoplakia) is, therefore, higher than that for histological ones. Because of their noninvasiveness and painlessness, cytological tests can also be used for mass screening. The most well-known and acknowledged cytologic screening is that for cervical cancer according to Papanicolaou. Screening of oral leukoplakias in patients with increased oral cancer risk, such as in Fanconi anemia, of sputum in patients with increased risk of lung cancer, and of urine in patients with increased risk of bladder cancer is well established. Applying adjuvant methods, such as immunocytochemistry, fluorescence in situ hybridization (FISH), or DNA cytometry, can help to increase diagnostic and typing accuracy [1]. In addition to the fact that most cancer cell-positive cytological diagnoses are subsequently validated by histological ones on tissues of the resected respective lesions, the diagnostic accuracy reaches high levels ^[2][3][4][5][2,3,4,5]. Furthermore, objective and prognostically valid grading of the malignancy of special types of cancer, such as the prostate, measuring the DNA content of thousands of cancer cells in enzymatic cell separation specimens, is another task that needs automation of diagnostic cytometry [6].

Cytopathology represents a subdiscipline of pathology and requires special diagnostic training. Screening cytological specimens needs high levels of concentration, in addition to being time-consuming and tiring. Specially trained cytotechnicians assist especially in screening repetitive smears, but the final cancer-cell-positive diagnoses must be validated by pathologists. Diagnostic accuracy greatly depends on the education and concentration of cytopathologists and cytotechnicians. It is, thus, very variable. At least in Germany, cytological investigations are less well paid by insurance companies and, accordingly, patients. This has a negative effect on their popularity among pathologists. Furthermore, the availability of skilled cytotechnicians is sinking, among others because most schools for their education have been closed in this country. In most countries of the third world, they are not available at all. Thus, the benefits of cytological screening for cancer cells and clarification of lesions suspicious for cancer do not reach most patients worldwide.

Thus, the availability of a computerized microscope which will be able to automatically screen cytological specimens from various sites of the human body for the presence of cancer cells and their precursors is very desirable. It should not replace cytotechnicians or cytopathologists but save their time and focus their sophisticated work on solving delicious diagnostic questions and confirming positive or suspicious diagnoses, e.g., proposed by the machine.

For about 25 years, automated Papanicolaou test screening devices have been commercially available. Two automated systems are in widespread use: the ThinPrep imaging system from Hologic (Marlborough, MA, USA) and the FocalPoint GS imaging system from Becton and Dickinson (Franklin Lakes, NJ, USA). They present a limited number of microscope fields of view with the highest likelihood of the presence of cellular abnormalities for further evaluation to cytotechnologists. These devices do not intend to detect single morphologically abnormal cells or to definitely identify cancer cells. They simply allow faster subjective screening [7].

2. Microscope-Based Scanners

The purpose of microscope-based scanners in the field of diagnostic cytology is to assist in screening repetitive cytological slides for the presence of cancer cells. They can identify and present regions of interest on a cytological slide that have to be assessed under a microscope in detail by skilled personnel, or they can definitively identify individual malignant or dysplastic cells (i.e., their respective nuclei). The latter solution will be more time-saving and will require less skilled personnel. We herein present such a solution. If nuclear DNA content, i.e., DNA aneuploidy, is used as a specific marker of malignant cells as diagnostic parameter, as in theour solution, a DNA-specific stain, such as that proposed by Robert Feulgen ^[8][21], is required. Thionine or pararosaniline can be applied. This requires a temperature (25 °C)-controlled acid (HCl)-resistant cuvette. The procedure can be performed automatically using acid-resistant staining machines comprising a temperature-controlled cuvette. Re-staining of samples according to Papanicolaou is possible for a subsequent subjective inspection of problematical slides. The microscope-based scanner should comprise a device for automated loading between 10 and 100 slides. The microscope should be equipped with a 40× high-NA objective. Scanning should be fast, completed within a few minutes per slide. A mechanical stepper should allow catching several images per field of view at different levels of focus as cells in cytological slides are mostly found at different z-levels. The color CMOS camera should allow high spatial and photometric resolutions (Figure 1). Diagnostic DNA karyometry represents the combination of automated morphometric classification of nuclei into those of different cell types, including reference cells, normal and suspicious nuclei, and DNA image cytometry. While the percentage of morphologically abnormal nuclei within a slide (e.g., >4%) can be used to raise suspicion of malignancy, the presence of DNA aneuploidy (single-cell and/or stemline aneuploidy) detected in the fraction of morphologically abnormal nuclei can be used as a specific marker of malignant cells. The combination of both features provides a means for a higher diagnostic accuracy. In theour series of 92 oral smears, DNA aneuploidy alone recognized only 40% of malignant clones; adding nuclear morphometry with the occurrence of >4% abnormal nuclei raised the sensitivity to 60%. Another task of diagnostic nuclear morphometric classification, in addition to the identification of reference cell nuclei, is to remove diagnostically irrelevant ones (e.g., lymphocytes, granulocytes, nuclear doublets, defocused nuclei, or artefacts) from diagnostic DNA cytometry. The universal algorithms suitable for diagnostic interpretation of nuclear DNA histograms have been internationally consented and published by the European Society for Analytical Cellular Pathology, ESACP ^{[9][10][11][12]}[26,27,28,30]. The percentage of abnormal nuclei suitable for raising a suspicion of malignancy depends on the type of tissue, the fixation and staining of specimens, and the nuclear classifier applied. For effusion specimens scanned with a MotiCyte-auto, rwesearchers found a suitable threshold of >0.75% [4]; for oral smears, reswearchers herein reported >4% to be useful. Cytotechnicians or cytopathologists can easily control the precision of nuclear classifications using image galleries and of diagnostic interpretations on DNA histograms. A cytopathological diagnosis should only be issued after a skilled operator has reviewed all nuclei on image galleries classified as morphologically abnormal and their respective DNA histogram in order to confirm DNA euploidy or aneuploidy.

3. Screening Oral Smears for Cancer Cells

Automated classification of normal and abnormal epithelial nuclei from oral smears (Figure 1, Figure 4 and Figure 5) amounted to 86.7% in contrast to 80% correctness after subjective control on image galleries. Using automatically classified and subjectively controlled in contrast to corrected nuclei of squamous epithelial cells as the internal reference to define the normal 2c value, controlled morphologically abnormal nuclei for analysis, and DNA aneuploidy as a marker of malignancy, a sensitivity of 44% was found for the detection of malignant nuclei according to theour modified follow-up. When adding the percentage of morphologically abnormal nuclei of 5% in contrast to 4% as an indicator of nuclei suspicious for malignancy, sensitivity raised to 64% in contrast to 72%. Respective specificities were 100% and 98.5% (Table 1b). DNA stemline and single-cell aneuploidy served as a highly (close to 100%) specific marker of malignancy. An increased percentage of morphologically abnormal nuclei served as a marker for suspected malignancy. Thus, the increased percentage of morphologically abnormal nuclei per slide helps to increase the diagnostic sensitivity of diagnostic DNA karyometry of oral smears. The reason for false-negative diagnoses most likely was not a paucity of cells but sampling errors, as eight of nine false-negative slides contained sufficient (>1500) normal epithelial nuclei (up to 8090). False positives, based on DNA aneuploidy as a marker, did not occur. The only false suspicious slide contained 7.8% morphologically abnormal nuclei. The sensitivity and specificity of cytological assessment of oral smears according to an updated Cochrane systematic review are 90% and 94%, respectively [5]. The sensitivity of subjective cytomorphological investigation of oral smears from 737 Fanconi-anemia patients was reported by Velleuer et al. (2020) [5] to be 97.7%, with a specificity of 84.5%. The combination with manual DNA cytometry yielded a sensitivity of 100% and a specificity of 92.2%. Thus, one advantage of the our current method is its high specificity. The estimated mean time needed for visual control and correction of nuclei automatedly classified as abnormal on an EasyScan-AI screen is 3 min, while a subjective microscopical screen of a liquid-based cytological specimen requires about 6 min. For that purpose, well-educated and experienced personnel are required (cytotechnicians) that are not available anywhere worldwide.

4. Grading the Malignancy of Prostate Cancer

It is well known that a significant percentage of screening-detected prostate cancers do not progress, even without therapy, thus being suitable for the conservative strategy of active surveillance (45.3% in ^[13][31]). They may be identified by their low histological grade of malignancy on core biopsies. Unfortunately, the subjective assessment of histological grades of malignancy of prostate cancers according to Gleason is insufficient (45.7% according to ^[14][32] Relying on this parameter in order to choose active surveillance instead of active therapy is, therefore, critical. In a previous study on 80 patients, researchwers could prove the superior prognostic validity of an automatically obtained DNA grade of malignancy (Figure 2) in early-stage prostate cancer patients [5]. During the follow-up period of 4.1 years, the probability to exclude a progression of an untreated, localized prostate cancer under active surveillance was 100% for objective DNA karyometry, but only 80.9% for the subjective microscopic Gleason score [6]. This means that patients with Gleason score 6 and 7 prostate cancers, who reveal an objectively assessed DNA grade 1 of malignancy, can safely rely on this conservative strategy. The time-consuming selection of fibroblast nuclei for internal reference of the normal 2 c DNA content and of thousands of cancer cell nuclei for analysis can now be automatically performed by a microscope-based scanner such as the EasyScan-AI, specially trained to identify these nuclear types and to derive internationally consented DNA grades of malignancy ^[15][16][18,20]. Realizing the known heterogeneity of prostate cancers, the fact that thousands of nuclei can be analyzed using such a device (researchwers found 4669 in the mean) additionally resulted in a better representativity of measurements as compared with manual ones. Low numbers of cytological samples so far investigated with the new technology of automated diagnostic/prognostic DNA karyometry limit the representativity of theour respective data. Further studies considering more representative patients in order to better substantiate diagnostic accuracies as compared to subjective diagnoses/prognoses are required.

5. Screening Effusions for Cancer Cells

Serous effusions are very frequent events in patients from internal medicine. Approximately 40% of these malignant cells can be found, very often unexpectedly ^[14][32]. In these cases, they are the first evidence of a malignant tumor. Thus, all serous effusions have to be microscopically investigated for the presence of cancer cells. Yet, not only is this time-consuming, but there are not enough skilled personnel worldwide to perform this task with sufficient diagnostic accuracy. The average sensitivity of subjective cytological screening of serous effusions without adjuvant methods is 58% [3]. Therefore, a device that is able to automatically scan cell sedimentation slides from effusions for the presence of cancer cells is very welcome. RWesearchers re recently published that the precursor of theour EasyScanAI, specifically trained to classify nuclei from different cell types for subsequent DNA measurements in effusions, is able to perform such screening [4]. ReIn this searchers tudy we investigated slides from 136 patients with known follow-up. While manual DNA cytometry yielded only 34 normal mesothelial cells in the mean as an internal reference per slide, automated DNA karyometry found 3734 and 68 more abnormal nuclei per slide. TheOur reported sensitivity of 76.4% at a specificity of 100%, therefore, allows the application of theour method in daily routine.