Improving nonalcoholic fatty liver disease classification performance with latent diffusion models.

Integrating deep learning with clinical expertise holds great potential for addressing healthcare challenges and empowering medical professionals with improved diagnostic tools. However, the need for annotated medical images is often an obstacle to leveraging the full power of machine learning models. Our research demonstrates that by combining synthetic images, generated using diffusion models, with real images, we can enhance nonalcoholic fatty liver disease (NAFLD) classification performance even in low-data regime settings. We evaluate the quality of the synthetic images by comparing two metrics: Inception Score (IS) and Fréchet Inception Distance (FID), computed on diffusion- and generative adversarial network (GAN)-generated images. Our results show superior performance for the diffusion-generated images, with a maximum IS score of 1.90 compared to 1.67 for GANs, and a minimum FID score of 69.45 compared to 100.05 for GANs. Utilizing a partially frozen CNN backbone (EfficientNet v1), our synthetic augmentation method achieves a maximum image-level ROC AUC of 0.904 on a NAFLD prediction task.

Vaginal Electrical Impedance Detects the Fertile Window in Healthy Women: a Pilot Study / La Impedancia Eléctrica Vaginal Detecta la Ventana Fértil en Mujeres Sanas: un Estudio Piloto

ABSTRACT This work analyzes the electrical impedance (EI) measurement of cervical mucus (CM) using a device to determine the fertile window. In this prospective and longitudinal study, fourteen healthy women aged 18 to 44 were enrolled to evaluate three menstrual cycles. EI was measured through a medical device inserted into the vagina for two minutes daily. Patients were monitored by urine luteinizing hormone (LH) strip, blood collection, and vaginal ultrasound to visualize the dominant follicle. Finally, the predictive EI capacity was validated by the receiver operating characteristic (ROC) of anovulatory vs. ovulatory impedances. The peak of LH was 35.7 (±4.5) mUI/ml and the dominant follicle size was 15.45 mm (±0.559). There were statistical differences in EI measurements between the follicular and luteal phases vs. the ovulation phase (p<0.0361 and p<0.0160). After data normalization, an area under the ROC curve (AUC) of 0.713 (P value= 0.0253), a Youden J index of 0.4545Ω, a sensitivity of 63.6%, and a specificity of 81.8% were found. Low EI in the ovulatory period belongs to the LH ovulatory peak and follicular release. EI can be used for ovulation monitoring, birth control, or promoting pregnancy as a safe and innocuous method.

RESUMEN Este trabajo analiza la medición de la impedancia eléctrica (IE) del moco cervical (MC) mediante un dispositivo para determinar la ventana fértil. En este estudio prospectivo y longitudinal, se incluyeron 14 mujeres sanas de 18 a 44 años para evaluar tres ciclos menstruales. La IE se midió a través de un dispositivo médico colocado en la vagina durante dos minutos diarios. Las pacientes fueron monitoreadas con una tira de hormona luteinizante (LH) en orina, recolección de sangre y ultrasonido vaginal para visualizar el folículo dominante. Finalmente, la capacidad predictiva de IE fue validada por la curva ROC (receiver operating characteristic) de impedancias anovulatorias vs. ovulatorias. El pico de LH fue de 35.7(±4.5) mUI/ml; el folículo de tamaño dominante fue de 15.45 mm (±0.559). Se encontraron diferencias estadísticas para la medición de la IE de las fases folicular y lútea versus la fase de ovulación (p<0.0361 y p<0.0160). Después de la normalización de los datos, se encontró un área bajo la curva ROC (AUC) de 0.713 (valor de P = 0.0253), un índice de Youden J de 0.4545 Ω, sensibilidad del 63.6 % y especificidad del 81.8 %. La IE baja en el período ovulatorio que pertenece al pico ovulatorio de LH y liberación folicular. La IE se puede utilizar para el control de la ovulación, el control de la natalidad o la promoción del embarazo como método seguro e inocuo.