By Krys Lee and Kaitlyn Wilkie
Technological innovations are changing what's possible with women's health. For healthcare providers entering the market for OB/GYN ultrasound or mammography for breast cancer screening, it's important to know how capabilities are changing and what distinguishes a must-have feature from a nice-to-have.
Emerging trends in obstetrics and gynecology ultrasound for women’s health are 3D/4D, beam steering technology, and auto calculation. Meanwhile, traditional 2D imaging is greatly assisted by the introduction of matrix probes/transducers. Transducers with multiple rows of crystals create a matrix, and this matrix of elements electronically fires to render multiple planes of imaging in real time. These images save time and decrease work-related musculoskeletal disorders (WRMSD) by reducing technologist wrist strain.
In addition, automated harmonics and volume rendering auto calculations are improving diagnostic ultrasound studies in obstetrics and gynecology. Acquisition relies on quality 2D images for rendering that can take 20 to 50 milliseconds, depending on volume. 3D images can be obtained freehand, with and without position sensing and automated acquisition. Optimal diagnostic ultrasound data using the 2D images is heavily technologist dependent. Real-time 3D imaging, deemed 4D, which is typically used in obstetrical ultrasound, incorporates technologist skill backed by technological advances that aid in detection of fetal facial, limb, and spinal malformations, as well as investigations into cardiac anomalies.
Automated calculation software is also utilizing 3D renderings in fetal biometry to calculate measurements of femur length, abdomen and head circumference, and bi-parietal diameter. Ultrasound machines’ ability to store image data assists in detection of morphologic and growth anomalies by monitoring fetal measurements.
3D imaging in gynecological ultrasound has become a standard in imaging the coronal plane. This allows for improved assessment of pelvic structures and organs, and volume rendering can assist in diagnosis of pathology that may have gone undetected in 2D imaging — where coronal imaging is rarely successful in obtaining diagnostic images of the uterine surface and fundus. 3D imaging in gynecological ultrasound aids in diagnosing structural anomalies, locating intrauterine contraceptive devices, assessment of fibroids, endometrial assessment, diagnosis and monitoring of polycystic and other ovarian syndromes, detection of possible malignant structures, and assessment of reproductive vascularity.
Mammography is considered the gold standard in breast cancer detection; ultrasound plays a key role along with MR in breast imaging. Ultrasound has been used to further image areas of the breast after a mammogram, as well as to investigate breast lesions, palpable lumps, and nipple discharge in the population that have yet to have a mammogram (women under 35, pregnant and lactating women, and men). This is typically accomplished using hand-held ultrasound 2D imaging with automated harmonics. While matrix transducers can be used with this imaging to improve visualization, they are technologist dependent — and reproducing optimal images may vary between technologists with a range of skill and experience. Automated breast ultrasound (ABUS) has been emerging in use to address this caveat. ABUS is technologist independent in imaging and images dense breast more efficiently, requiring the technologist to properly position the patient. Meanwhile, the ultrasound scanner, a stationary device with a transducer, acquires images in the scan box. ABUS images may incorporate computer-aided detection (CAD) for improved 2D images that can be converted into multiplanar 3D images after reconstruction. This, along with a larger field of view, assists radiologists in detecting, identifying, and treating breast lesions.
To aid in detecting cancerous lesions in dense breasts, mammography system manufacturers have focused their attention on software solutions to improve image clarity and analysis of digital breast tomosynthesis (DBT) images in recent years. Deep learning artificial intelligence (AI) software continues to be of great interest in the mammography market. Deep learning AI outperforms conventional CAD software, which is limited by its programming; the algorithms must be set to identify characteristics of suspicious lesions, an activity prone to the subjectivity of the programmer. Any improvements to the software require a revision by the programmer. In contrast, deep learning AI software algorithms, which are made of multiple layers of neural networks, are capable of finding and scoring irregularities on their own, with performance continuously improving as the algorithms see more cases and more variable data sets.
AI software in mammography is currently being used to triage cases by irregularities found, and to aid the radiologist in interpreting images. The radiologist remains the final decision-maker on callbacks for additional diagnostics, as available AI software does not yet consistently outperform an experienced radiologist in a stand-alone setting. However, multiple studies have found using AI in conjunction with a radiologist’s visual interpretation improves cancer detection while decreasing false positives and unnecessary callbacks. AI software’s capabilities are also beneficial to clinics’ workflows. In helping radiologists prioritize cases and identify suspicious lesions, AI decreases reading time required on low-score images and overall workload. There is room in the women’s health market for deep learning AI software to grow in coming years, as demand for imaging services is outgrowing the percentage of practicing radiologists in the US.
About the authors: Kaitlyn Wilkie and Krys Lee are clinical advisors at TractManager, now a part of symplr.