Table 1C.

AI-based solutions for treatment efficacy prediction.

Studies	Cohorts (n)	Tumor localization	Treatment	Imaging modality	Models	Outcome prediction	Results	Reference
Clinical	124 patients	BM	Stereotactic radiation therapy (SRT)	MRI	MLP/Clinical features CNN + Seq2Seq/Transformers/LSTM CNN + Seq2Seq/Transformers/LSTM + clinical features	Local tumor control	CNN + LSTM + clinical features (AUC = 0.86, Acc = 0.83, Sens = 0.77, Spe = 0.87)	(36)
Clinical	30 patients	Gliomas (15 GBM)	/	MRI	HD-GLIO-XNAT (https://github.com/NeuroAI-HD/HD-GLIO-XNAT)	Evaluate whether AI-assisted decision support provides a more reproducible and standardized assessment of response to treatment compared to manual measurements using RANO criteria	Lower-grade gliomas (CCP = 0.77 for RANO and 0.91 with AI)	(37)
Clinical	133 patients	GBM	/	MRI	ANN with clinical features	Survival classification	Cross validation: Acc = 0.91	(38)
Preclinical	28 mice	GL261	Chemotherapy (Temozolomide)	MRI/MRS	1D-CNN, LR, SVM, RF, XGBoost	Therapy response assessment	1D-CNN (Acc = 0.9975, Sens = 0.99, Spe = 0.99)	(39)

Acc, accuracy; AI, artificial intelligence; AUC, area under the ROC curve; BM, brain metastasis; CCP, concordance correlation coefficients; CNN, convolutional neural network; GBM, glioblastoma; LR, logistic regression; LSTM, long short-term memory; MLP, multilayer perceptron; MRI, magnetic resonance imaging; MRS, magnetic resonance spectroscopy; RANO, response assessment in neuro-oncology; RF, random forest; Sens, sensitivity; Spe, specificity; SRS, stereotactic radiosurgery; SVM, support vector machine; XGBoost, extreme gradient boosting.