Automated Detection of Malignant Epithelial Cells in Effusions by DNA Karyometry
0Automated Detection of Malignant Epithelial Cells in Effusions by DNA-Karyometry
Layman Abstract : Background:
When doctors check fluid samples (like from the lungs or abdomen) to look for cancer cells, the usual method — called conventional cytology — only finds cancer about 58% of the time, meaning it misses many cases. Another method, called DNA-Image-Cytometry (DNA-ICM), which measures the amount of DNA in suspicious cells using images, can find cancer cells in up to 91% of cases. However, the downside is that this process takes a long time — up to an hour to check just one slide.
Aim:
This study tested a new, faster, computer-assisted method called DNA karyometry (DNA-KM). It helps detect cancer cells in fluid samples more quickly and efficiently, reducing the expert’s workload.
Methods:
The researchers developed a computer program using machine learning (a type of artificial intelligence) to automatically identify suspicious cells in stained fluid samples. They tested this program on 121 patient samples and compared its accuracy to both manual DNA-ICM (where a person does all the work) and traditional cytology (the usual method doctors use).
Results:
The computer program was much faster — it found about 100 times more cells per slide than a person could find manually (average of 3734 cells vs. 34 cells). It also found more abnormal (possibly cancerous) cells than the manual method.
The new method was also more sensitive, meaning it caught more cases of cancer than manual DNA-ICM:
New method sensitivity: 76.4%
Manual method sensitivity: 68.5%
Both methods had 100% specificity, meaning when they said cancer was present, they were always correct.
The system also created a helpful new marker, the ratio of abnormal cells to total cells, which could correctly identify all cancer cases (100% sensitivity) but with 70% specificity (meaning 30% of the time it might falsely suggest cancer is present).
Finally, the new method reduced the expert’s work time from 60 minutes per slide to just 5 minutes, since the computer does most of the work, and the expert only has to double-check a few suspicious cells.
Conclusion:
This new computer-assisted DNA karyometry method makes cancer detection in fluid samples faster, easier, and more accurate compared to the current manual method. This is especially important in many parts of the world where there aren’t enough specialists to check all the samples. By combining computer automation with expert review, this approach could improve cancer detection and save time.
----------
Original Abstract : Background: The average sensitivity of conventional cytology for the identification of malignant epithelial cells in effusions is only about 58%. DNA-Image-Cytometry (DNA-ICM), which exploits the DNA content of morphologically suspicious nuclei measured on digital images, has a sensitivity for the detection of cells of up to 91%. Yet, when performed manually, an expert so far needs about 60 minutes for the analysis of a single slide.
Aim: This study presented and evaluated a novel solution for rapid, computer-assisted, semiautomated diagnostic DNA cytometry of serous effusion specimens: DNA karyometry (DNA-KM).
Methods: A novel method of supervised machine learning is presented for the automated identification of morphologically suspicious mesothelial and epithelial nuclei in Feulgen-stained effusions. This method was compared to manual DNA-ICM and a gold-standard cytological diagnosis for 121 cases. Furthermore, the potential of using the amount of morphometrically abnormal mesothelial or epithelial nuclei detected by the digital classifier as an additional diagnostic marker was analyzed. SPSS statistical software (version 22.0.001, IBM Corporation, Armonk, New York) was used.
Results: The mean number of lymphocytes automatically identified per slide by digital nuclear classifiers was approximately 100 times higher than the number selected manually (3734.1 vs 33.8). The presented semi-automated DNA-karyometric solution identified more diagnostically relevant abnormal nuclei than manual DNA-ICM, which led to a higher sensitivity (76.4 vs. 68.5%) at 100% specificity. The ratio between digitally abnormal and all mesothelial nuclei can identify cancer-cell positive slides at 100% sensitivity and 70% specificity. The time-effort for an expert is thus reduced to the morphological verification of a few nuclei with exceeding DNA-content, which can be accomplished within five minutes........See More: https://doi.org/10.9734/bpi/msti/v8/4556
#DNA_Karyometry #DNA_image_cytometry #serous_effusions #nuclear_classifiers #automated_cytology