|
|
Albiol, F., Corbi, A., & Albiol, A. (2019). Densitometric Radiographic Imaging With Contour Sensors. IEEE Access, 7, 18902–18914.
Abstract: We present the technical/physical foundations of a new imaging technique that combines ordinary radiographic information (generated by conventional X-ray settings) with the patient's volume to derive densitometric images. Traditionally, these images provide quantitative information about tissues densities. In our approach, they graphically enhance either soft or bony regions. After measuring the patient's volume with contour recognition devices, the physical traversed lengths within it (as the Roentgen beam intersects the patient) are calculated and pixel-wise associated with the original radiograph (X). In order to derive this map of lengths (L), the camera equations of the X-ray system and the contour sensor are determined. The patient's surface is also translated to the point-of-view of the X-ray beam and all its entrance/exit points are sought with the help of ray-casting methods. The derived L is applied to X as a physical operation (subtraction), obtaining soft tissue-(D-S) or bone-enhanced (D'(B)) figures. In the D-S type, the contained graphical information can be linearly mapped to the average electronic density (traversed by the X-ray beam). This feature represents an interesting proof-of-concept of associating density data to radiographs, but most important, their intensity histogram is objectively compressed, i.e., the dynamic range is more shrunk (compared against the corresponding X). This leads to other advantages: improvement in the visibility of border/edge areas (high gradient), extended manual window level/width manipulations during screening, and immediate correction of underexposed X instances. In the D-B' type, high-density elements are highlighted and easier to discern. All these results can be achieved with low-energy beam exposures, saving costs and dose. Future work will deepen this clinical side of our research. In contrast with other image-based modifiers, the proposed method is grounded on the measurement of a physical entity: the span of the X-ray beam within a body while undertaking a radiographic examination.
|
|
|
|
Schaffter, T. et al, Albiol, F., & Caballero, L. (2020). Evaluation of Combined Artificial Intelligence and Radiologist Assessment to Interpret Screening Mammograms. JAMA Netw. Open, 3(3), e200265–15pp.
Abstract: Importance Mammography screening currently relies on subjective human interpretation. Artificial intelligence (AI) advances could be used to increase mammography screening accuracy by reducing missed cancers and false positives. Objective To evaluate whether AI can overcome human mammography interpretation limitations with a rigorous, unbiased evaluation of machine learning algorithms. Design, Setting, and Participants In this diagnostic accuracy study conducted between September 2016 and November 2017, an international, crowdsourced challenge was hosted to foster AI algorithm development focused on interpreting screening mammography. More than 1100 participants comprising 126 teams from 44 countries participated. Analysis began November 18, 2016. Main Outcomes and Measurements Algorithms used images alone (challenge 1) or combined images, previous examinations (if available), and clinical and demographic risk factor data (challenge 2) and output a score that translated to cancer yes/no within 12 months. Algorithm accuracy for breast cancer detection was evaluated using area under the curve and algorithm specificity compared with radiologists' specificity with radiologists' sensitivity set at 85.9% (United States) and 83.9% (Sweden). An ensemble method aggregating top-performing AI algorithms and radiologists' recall assessment was developed and evaluated. Results Overall, 144231 screening mammograms from 85580 US women (952 cancer positive <= 12 months from screening) were used for algorithm training and validation. A second independent validation cohort included 166578 examinations from 68008 Swedish women (780 cancer positive). The top-performing algorithm achieved an area under the curve of 0.858 (United States) and 0.903 (Sweden) and 66.2% (United States) and 81.2% (Sweden) specificity at the radiologists' sensitivity, lower than community-practice radiologists' specificity of 90.5% (United States) and 98.5% (Sweden). Combining top-performing algorithms and US radiologist assessments resulted in a higher area under the curve of 0.942 and achieved a significantly improved specificity (92.0%) at the same sensitivity. Conclusions and Relevance While no single AI algorithm outperformed radiologists, an ensemble of AI algorithms combined with radiologist assessment in a single-reader screening environment improved overall accuracy. This study underscores the potential of using machine learning methods for enhancing mammography screening interpretation. Question How do deep learning algorithms perform compared with radiologists in screening mammography interpretation? Findings In this diagnostic accuracy study using 144231 screening mammograms from 85580 women from the United States and 166578 screening mammograms from 68008 women from Sweden, no single artificial intelligence algorithm outperformed US community radiologist benchmarks; including clinical data and prior mammograms did not improve artificial intelligence performance. However, combining best-performing artificial intelligence algorithms with single-radiologist assessment demonstrated increased specificity. Meaning Integrating artificial intelligence to mammography interpretation in single-radiologist settings could yield significant performance improvements, with the potential to reduce health care system expenditures and address resource scarcity experienced in population-based screening programs. This diagnostic accuracy study evaluates whether artificial intelligence can overcome human mammography interpretation limits with a rigorous, unbiased evaluation of machine learning algorithms.
|
|
